6G-IoT coherent feature GAT fusion suppression superposition CSI feedback method

The invention discloses a 6G-IoT coherent feature GAT fusion suppression superposition CSI feedback method, which comprises the following steps that: a user side extracts downlink CSI coherent features and feeds back downlink CSI to a base station side by adopting a spread spectrum superposition CSI feedback mode; the base station end carries out equalization processing on the received signal to obtain a superposition sequence initial feature; and according to the superposition sequence initial features, constructing a multi-task superposition interference suppression network based on GAT and a fusion interference suppression mechanism, and obtaining feedback downlink CSI reconstruction features and uplink user data sequence detection features. Compared with non-superposed CSI feedback, the method has the advantages that the energy consumption of the user side and the data transmission resource of the 6G-IoT system are saved; compared with superposed CSI feedback, according to development ...

Method and system for determining recommendation information, computing equipment and storage medium

The invention discloses a method and system for determining recommendation information, computing equipment and a storage medium. The method for determining the recommended information comprises the steps of constructing a directed graph of to-be-recommended information based on historical behaviors of a user; based on the directed graph, using random walk to generate a plurality of sequences for the to-be-recommended information; the features of the to-be-recommended information are processed through a vector generation model to generate feature vectors of the to-be-recommended information, and training samples of the training vector generation model are generated by the sequences; weighting the feature vector of each piece of to-be-recommended information pointed by the historical behavior of the user to generate a user feature vector; and in response to a recommendation information acquisition instruction, determining recommendation information based on the similarity between the user ...

Method and system for mining and predicting spatial-temporal trajectory of mobile user

The invention provides a method and a system for mining and predicting a spatio-temporal trajectory of a mobile user, and the scheme comprises the steps: carrying out the preprocessing of the obtained historical spatio-temporal trajectory data of the current mobile user, and obtaining a trajectory sequence sorted according to the time and a spatio-temporal context attribute set; carrying out feature embedding on the track sequence and the user to respectively obtain track sequence features and user features; inputting the feature-embedded trajectory sequence into a hierarchical attention architecture using a pre-trained self-attention neural network as a backbone network to obtain long-term trajectory features and short-term trajectory features; based on the embedded user features, performing feature fusion on the long-term trajectory features and the short-term trajectory features to obtain fused trajectory features; and based on the fused trajectory features, processing by using a full ...

Early diagnosis method for abnormity of power module of DCS (Distributed Control System) device

A DCS device power supply module abnormity early diagnosis method comprises the following steps: obtaining an output ripple signal generated by a monitored power supply module in a preset time period, and load values of the monitored power supply module at a plurality of preset time points in the preset time period; an artificial intelligence technology based on deep learning is adopted, a deep learning technology is utilized, features are extracted from an output ripple signal and a load value of a switching power supply, and fault diagnosis of a monitored power supply module is realized by utilizing classification processing. Therefore, the switching power supply can be effectively monitored, the abnormal state of the power supply can be found in time, and the reliability and the safety of the power supply are improved.

Geological safety monitoring method and device, computer equipment and storage medium

The invention relates to a geological safety monitoring method and device, computer equipment, a storage medium and a computer program product. The method comprises the following steps: acquiring a historical abnormal state information set, a geographic state information set, an environment state information set and a policy state information set corresponding to a target monitoring area; fusing each piece of historical abnormal state information in the historical abnormal state information set to obtain a historical factor, and obtaining a geographic factor based on each piece of geographic state information in the geographic state information set; fusing the historical factor and the geographic factor to obtain a historical early warning value; fusing each piece of environment state information in the environment state information set to obtain an environment factor, and obtaining a policy factor based on each piece of policy state information in the policy state information set; fusing ...

Microseismic event identification method and system based on multi-modal neural network

The invention discloses a microseism event identification method and system based on a multi-modal neural network, and the method comprises the following steps: obtaining a time domain sample and an S domain sample, and constructing a training data set based on the time domain sample and the S domain sample; constructing a deep learning network model G, and setting a model loss function; training the deep learning network model G based on the training data set to obtain an identification model; and carrying out microseism event identification by using the identification model. According to the method, the deep learning convolutional neural network structure of the time domain mode and the S domain mode of the microseismic data is combined, the multi-mode information of the microseismic signals is fully excavated, the accuracy of the model is improved, the method can still have good performance under strong noise interference, events can still be well recognized, and the method is suitable ...

Project recommendation method and device, equipment and storage medium

The invention relates to artificial intelligence and digital medical treatment, and provides a project recommendation method and device, equipment and a storage medium. The method comprises the following steps: generating an interaction matrix based on interaction information of an initial user to an initial project, generating a user walk sequence and a project walk sequence based on the interaction matrix, inputting the user walk sequence into a user representation model to obtain a user representation vector, processing the project walk sequence based on the project representation model to obtain a project representation vector, and displaying the project representation vector. Generating a user context vector based on the multi-dimensional information of the target user, generating an item context vector based on the multi-dimensional feature of each initial item, and generating an interested item vector based on the item representation vector and the item context vector, the target ...

Urban traffic prediction method and device based on artificial intelligence, and medium

The invention discloses an urban traffic prediction method and device based on artificial intelligence, and a medium, and the method comprises the steps: constructing a graph network containing traffic flow, average vehicle speed, and average lane space occupancy information, employing the CGN graph convolution to extract the spatial features of traffic flow, adding a long memory filter in an RNN network, extracting the local time feature information, and carrying out the prediction of the traffic flow. Global time features are obtained through a multi-head attention mechanism, and the local time and the global time features are fused and output, so that efficient prediction of future traffic conditions is realized.

Human body tumble detection method and system based on millimeter wave radar and deep learning

The invention belongs to the field of human body tumble detection, and discloses a human body tumble detection method and system based on a millimeter wave radar and deep learning, and the method comprises the following steps: obtaining an echo signal of the millimeter wave radar; performing short-time Fourier transform on the echo signal to obtain a time-frequency diagram; extracting point cloud data in the echo signal, and extracting human body point cloud cluster features from the point cloud data; and inputting the point cloud feature data and the time-frequency graph into an LSTM network subjected to block item decomposition, carrying out feature fusion through a full connection layer, and finally carrying out classification by using a softmax layer to complete human body tumble detection and recognition training. And inputting the features of the point cloud cluster group obtained in real time and the time-frequency graph into a pre-trained LSTM network, and judging whether the old ...

Information fusion algorithm based on target track data element weights

The invention discloses an information fusion algorithm based on target track data element weights, and relates to the technical field of information processing, and the algorithm comprises the steps: taking the time generated by radar target data as a time dimension, so as to generate a plurality of pieces of single-point data; single-point data is made by adopting element weight scores; carrying out fusion comparison on the radar target track point and the AIS target track point obtained at the same moment, and if the fusion score is greater than a set threshold value, determining that the target point is the same target point; the data fusion terminal collects N points for flight path fusion judgment; when the proportion of the data points which are true in the fusion validity period is greater than a preset value, the radar target and the AIS target are fused into a target track, so that the track association accuracy is improved; performing track fusion monitoring on the data fusion ...

Model training method and device, electronic equipment and storage medium

The invention provides a model training method and device, electronic equipment and a storage medium. Comprising the steps of obtaining a multi-modal model training sample; inputting the multi-modal model training sample into a to-be-trained multi-modal sentiment analysis model; calling a vector representation network layer to process the text modal feature, the voice modal feature and the video modal feature to obtain a text representation vector, a voice representation vector and a video representation vector; calling a multi-modal fusion network layer to perform multi-modal fusion processing on the text representation vector, the voice representation vector and the video representation vector to obtain a multi-modal fusion feature vector; calling a sentiment classification network layer to process the multi-modal fusion feature vector to obtain a classification probability of a sentiment category to which a multi-modal model training sample belongs; calculating a loss value of the to-be-trained ...

Federal domain adaptive fault diagnosis method based on statistical feature fusion

The invention discloses a federated domain adaptive fault diagnosis method based on statistical feature fusion. A source domain client receives statistical features of a target domain client and statistical features of self-active source domain data and performs data standardization on a source domain data set; the central server sends the fault diagnosis model to all source domain clients, and then the source domain clients use two kinds of standardized source domain data as input to train a received global model in the source domain clients; the trained model is returned to the central server, and the central server carries out global model aggregation on the received models of all the source domain clients; when the training round number reaches a set value, training is finished, a final fault diagnosis model is obtained and sent to the target domain client side, the target domain client side obtains mechanical fault data to be diagnosed, and a fault diagnosis result is obtained through ...

Single-lead electrocardiosignal auxiliary diagnosis method and diagnosis terminal

The invention provides a single-lead electrocardiosignal auxiliary diagnosis method and a diagnosis terminal, and belongs to the field of internet medical treatment, the method comprises the following steps: constructing a functional architecture of a single-lead electrocardiosignal auxiliary system, and extracting electrocardiosignals in the single-lead electrocardiosignal auxiliary system; the electrocardiosignals are preprocessed; the preprocessing mode comprises the steps that the electrocardiosignals are divided according to preset duration; carrying out filtering processing on the divided electrocardiosignals; carrying out data normalization processing on the filtered electrocardiosignal; fourier transform processing is carried out on the electrocardiosignal after data normalization; the characteristics of the electrocardiosignal data are improved; fusing all the extracted features in a splicing mode; and inputting the fused features into an Adaboost iterative algorithm to obtain ...

Text similarity analysis method and system based on residual fusion and stack integration

The invention discloses a text similarity analysis method and system based on residual fusion and stack integration. A text similarity analysis method based on residual fusion and stack integration comprises the steps that a to-be-analyzed text is input into a target text similarity judgment model for similarity judgment, and a judgment result is obtained; the target text similarity judgment model is obtained by fusing a representation type SBERT model and an interactive type ABCNN model; if the to-be-analyzed text is judged to be the similar text, inputting the to-be-analyzed text into a target text similarity calculation model for similarity calculation to obtain a target text similarity value; the target text similarity calculation model is obtained by performing parallel heterogeneous integration on a plurality of different interactive pre-training models. The technical problem that an existing text similarity analysis model is difficult to have high retrieval efficiency and accuracy ...

Text error correction method and device, electronic equipment and storage medium

The invention relates to the technical field of language processing, further relates to the technical field of artificial intelligence, and particularly relates to a text error correction method and device, electronic equipment and a storage medium. A to-be-processed text is acquired, and the to-be-processed text is input into a feature coding model to obtain a first feature vector; inputting the first feature vector into a text error type classification model, and outputting a first probability value, a second probability value and a third probability value; inputting the first feature vector into a first graph convolutional neural network model based on a homophone graph network to obtain a second feature vector; inputting the first feature vector into a second graph convolutional neural network model based on a similar word graph network to obtain a third feature vector; weighting the first feature vector, the second feature vector and the third feature vector to obtain a fused feature ...

Knowledge graph completion method fusing text description and graph convolution mechanism

The invention provides a knowledge graph completion method fusing text description and a graph convolution mechanism, and belongs to the field of knowledge graph completion. The method sequentially comprises the following steps: firstly, taking an entity description text and a relational word as input, and acquiring information from a neighborhood by utilizing a multilayer weighted graph convolutional network, integrating entity semantics and relational word vectors and an attention mechanism of entity semantics and adjacent entity semantics to obtain high-quality entity features; secondly, through cross-modal Transfomer and cavity convolution, interaction between the entity and the relationship is further improved, and a feature map is obtained; finally, the feature graph is leveled and then mapped to the space where the entity vector is located, and the similarity score is obtained after point multiplication with the tail entity. According to the method, better entity embedding representation ...

Multi-modal data fusion analysis method for Alzheimer's disease pathogenesis

The invention provides a multi-modal data fusion analysis method for Alzheimer's disease pathogenesis. The multi-modal data fusion analysis method comprises the following steps: acquiring functional magnetic resonance imaging image data of a patient and single nucleotide polymorphism of a gene for preprocessing; carrying out matrix fusion on the processed image and a gene and then constructing a brain region gene pair; and performing feature extraction and model establishment on the brain region gene pair through an improved multi-core support vector machine. According to the multi-modal data fusion analysis method for the Alzheimer's disease pathogenesis, the thought of a prior algorithm is applied to a multi-core SVM, the advantages of the two algorithms are utilized, feature combinations are dynamically constructed, classification results are continuously verified, all data are verified in sequence, random occasion is avoided, and it is guaranteed that each step of an experiment is guaranteed ...

Data classification method and device, electronic equipment and storage medium

The invention provides a data classification method and device, electronic equipment and a storage medium. The method comprises the steps of obtaining to-be-classified data; classifying the to-be-classified data by adopting a target classifier based on an evidence reasoning rule to obtain a classification result; the training of the target classifier comprises obtaining a data set; performing evidence acquisition on the samples in the training set; constructing an optimization objective function according to the degree matrix, and optimizing the feature reference value through the optimization objective function; determining the sample number of the sample category according to the features, and determining the reliability of the evidence; performing evidence activation and evidence combination processing on the test set and the verification set through the reliability matrix; and calculating a target function value of the optimization target function according to the category prediction ...

Service recommendation method based on depth matrix decomposition and knowledge graph

The invention relates to a service recommendation method based on depth matrix decomposition and a knowledge graph, and the method comprises the steps: constructing a government affair service knowledge graph through government affair service item data, and carrying out the modeling of entity context and entity description text information of the knowledge graph through a knowledge representation method; knowledge representation learning and personalized recommendation are combined by adopting a joint learning mode to obtain an optimal GKGR model. And finally, predicting scores of each pair of users and service items, and recommending the service items with higher scores to the users in a list recommendation mode. According to the method, the neural network is utilized to perform feature extraction on users and service items, user behavior data are fully utilized, and the problem of data sparsity is effectively relieved; the entity context and the entity description text information are ...

Robust visual question and answer model training method based on comparative learning

The invention discloses a robust visual question and answer model training method based on comparative learning, and belongs to the technical field of natural language processing and computer vision cross application. In the aspect of image enhancement, an enhancement method based on visual context disturbance is used, a visual context weak in correlation with a problem is screened out through attention distribution of objects in an image, disturbance is added to the visual context to construct new image representation, and therefore a model learns image representation irrelevant to the visual context. Performing text enhancement on questions of general interrogative sentence types by adopting a strategy of deleting interrogative assistant words; for other types of problems, a rewriting strategy is adopted for text enhancement. A positive sample is constructed through the data enhancement method, optimization is performed by using a comparative learning method, and unbiased multi-modal ...

Visual information management method and system for site pollution data and medium

The invention discloses a visual information management method and system for site pollution data and a medium, and the method comprises the steps: obtaining multi-source heterogeneous data of a target pollution site, and constructing a three-dimensional visual model of the target pollution site; performing data fusion by using a graph convolutional neural network to obtain space-time correlation features, and extracting pollutant space-time distribution of the current time step length; semi-supervised learning is carried out according to historical multi-source monitoring data of the target pollution site to obtain a risk information early warning model, and correlation between pollution characteristics and soil and groundwater environment quality is obtained; the spatial and temporal distribution of pollutants is imported into the risk information early warning model, pollution risk identification is carried out, and pollution risk early warning and soil and groundwater environment quality ...

Lightweight convolutional neural network and Transform fusion-based arrhythmia detection method

The invention provides an arrhythmia detection method based on lightweight convolutional neural network and Transform fusion, and relates to the technical field of ECG classification. The method comprises the steps that firstly, electrocardiosignals are collected, and preprocessing including resampling, segmentation and normalization is conducted on electrocardiosignal data; generating an electrocardiosignal training data set, and dividing the electrocardiosignal training data set into a training set and a test set; the method comprises the following steps of: constructing a trunk CNN (Convolutional Neural Network) by utilizing a sandglass block, and introducing hyper-parameters to construct an arrhythmia feature extraction network; constructing an arrhythmia classification network by using a Transform model and a feed-forward neural network (FFN); the performance of the model is evaluated, the electrocardiosignal training set is input into the arrhythmia detection network, and the arrhythmia ...

Multi-modal named entity recognition method based on strong interaction and fine-grained alignment

The invention relates to a multi-modal entity recognition method in the field of natural language processing, in particular to a method for obtaining more accurate information correlation among models aiming at insufficient information interaction among modals, and the method comprises the following steps: segmenting a text according to words, converting the segmented text into digital marks, inputting the digital marks into a BERT pre-training model, and obtaining a vector of a last hidden layer; inputting corresponding image information into the RESNET model to obtain an image hidden layer vector, identifying a target contained in the image by using a target detection tool, and judging the correlation between modals by calculating the correlation between an entity and the target in the image; the embedding distribution distance of the text vector and the image vector is shortened in a comparative learning mode, and the representation significance of the text vector is optimized; meanwhile ...

Mechanical ventilation man-machine asynchronous single-stage identification method and system

The invention discloses a mechanical ventilation man-machine asynchronous single-stage identification method and system. The method comprises the steps that pressure time sequence, flow speed time sequence and volume time sequence data of a breathing machine are obtained; after feature embedding and feature extraction are carried out on the pressure time sequence data, the flow velocity time sequence data and the volume time sequence data, feature maps representing the pressure time sequence data, the flow velocity time sequence data and the volume time sequence data are obtained; an output tensor is obtained after the characteristic spectrum passes through a classifier, and the position where man-machine asynchronization occurs and the type of man-machine asynchronization are obtained based on the output tensor. According to the mechanical ventilation man-machine asynchronization single-stage identification method, positioning and classification can be fused in one characteristic spectrum ...

Multi-sensor fusion quadrotor unmanned aerial vehicle attitude estimation method

The invention provides a multi-sensor fusion quadrotor unmanned aerial vehicle attitude estimation method. The method comprises the following steps: S1, establishing a state observation model of a quadrotor unmanned aerial vehicle; s2, filtering processing is carried out on the sensor signals of the multiple channels; s3, linear combination parameters of the multi-channel sensor signals are calculated; s4, calculating a sensor value in a fixed axial direction according to the linear combination parameters; and S5, constructing an inverse solution equation of the observation model, and performing attitude solution according to an observation result. The method has the advantages of being high in operation speed, good in measurement stability and high in response speed.

Electroencephalogram signal analysis method based on multi-band fusion and space-time Transform

The invention discloses an electroencephalogram signal analysis method based on multi-band fusion and space-time Transform, which comprises the following steps: firstly, converting a preprocessed electroencephalogram signal into a multi-band image with a corresponding time length, and reserving three-dimensional space information between sampling channels as far as possible; then fusing features by adopting a frequency band attention module for calculating an attention map of the stacked multi-frequency-band images and deducing a fused feature map; using a time/space self-attention module to extract spatial and temporal features for representing and distinguishing multi-frame and multi-region dynamic brain states; and finally, learning category information of the features through a multi-layer perceptron. The MEET model obtained through training can represent and analyze a multi-scale time sequence of the human brain electroencephalogram signals. According to the method, the accuracy of ...

Circuit breaker operation state diagnosis method and system and readable computer storage medium

The invention discloses a circuit breaker operation state diagnosis method, which comprises the following steps of: installing a plurality of sensors on a circuit breaker, extracting the characteristics of a sampling signal of a vibration sensor by utilizing a wavelet noise reduction theory and a wavelet packet decomposition algorithm, and extracting the characteristics of a circuit breaker stroke-time characteristic signal acquired by a displacement sensor; and determining the membership degree of the monitored circuit breaker belonging to different operation states in combination with the typical feature vectors stored in the fault feature library. And the credibility of each sensor is introduced as a weight, so that a weighted membership degree is obtained, and finally, a decision result, namely the current operation state of the circuit breaker machinery, is obtained. According to the circuit breaker state diagnosis method, fusion of information collected by multiple sensors is achieved ...

User abnormal behavior characteristic analysis system based on deep learning and use method

The invention provides a user abnormal behavior characteristic analysis system based on deep learning and a use method. The user abnormal behavior characteristic analysis system comprises a characteristic extraction module, a time characteristic analysis module, a space characteristic analysis module and an integration anomaly analysis module. The feature extraction module is used for extracting features according to user behaviors; the time characteristic analysis module is used for splicing the marked basic characteristics and time measurement indexes of the basic characteristics; the spatial characteristic analysis module is used for acquiring marked basic characteristics of users of the same role and training a spatial sharing group model; the integration anomaly analysis module is used for reconstructing an error according to a time sample obtained by the time representation model and a space sample obtained by the space sharing group model; by applying the technical scheme, synchronous ...

State evaluation method and device of intelligent high-voltage switch, terminal and storage medium

The invention provides a state evaluation method and device of an intelligent high-voltage switch, a terminal and a storage medium. The method comprises the following steps: acquiring a sensing data set of the intelligent high-voltage switch; respectively training a stack denoising auto-encoder SDAE and a deep belief network DBN according to the sensing data set to obtain the weight of the SDAE and the weight of the DBN; fusing the weight of the SDAE and the weight of the DBN by adopting a weighted fusion method to obtain a fused weight, initializing the weight of the target evaluation model according to the fused weight, and testing the feasibility of the target evaluation model; and when the feasibility of the target evaluation model meets a preset condition, obtaining operation data of the intelligent high-voltage switch, and inputting the operation data into the target evaluation model to obtain an operation state of the operation data. The method can accurately and efficiently evaluate ...

Solar panel fault detection method based on power data and visual analysis

The invention discloses a solar panel fault detection method based on power data and visual analysis, belongs to the technical field of solar cells, and solves the problem that the existing solar panel fault detection method cannot accurately analyze and judge the fault of a solar panel in combination with the operation power of the solar panel and visual imaging. The method comprises the following steps: acquiring real-time operation power data of a solar panel; obtaining a power anomaly feature set; acquiring a real-time visual imaging set of the solar panel, and performing fusion analysis and diagnosis based on the fault fusion diagnosis model to obtain a solar panel fault diagnosis result; the solar panel is monitored and early warned through the composite fault fusion diagnosis model, the generation power abnormity of the solar panel can be found in time, the generation power of the solar panel and the real-time visual imaging set can be combined, the generation abnormal point of the ...

SOC estimation method and device of energy storage system, equipment and storage medium

The embodiment of the invention discloses an SOC (State of Charge) estimation method and device for an energy storage system, equipment and a storage medium, and the method comprises the steps: carrying out the SOC estimation of the energy storage system based on the charging and discharging characteristics of the energy storage system and the collected current data, voltage data and temperature data of the energy storage system; determining an interaction model of a high-frequency equivalent circuit and an interaction model of a low-and-medium-frequency equivalent circuit of the energy storage system, wherein the charging and discharging characteristics at least comprise charging and discharging hysteresis characteristics of the energy storage system; particle filtering processing is carried out on the interaction models, and the estimated SOC of each interaction model is determined; and performing SOC fusion processing by using the estimated SOC of each interactive model and a preset ...

Gesture recognition using multiple antenna

Various embodiments wirelessly detect micro gestures using multiple antenna of a gesture sensor device. At times, the gesture sensor device transmits multiple outgoing radio frequency (RF) signals, each outgoing RF signal transmitted via a respective antenna of the gesture sensor device. The outgoing RF signals are configured to help capture information that can be used to identify micro-gestures performed by a hand. The gesture sensor device captures incoming RF signals generated by the outgoing RF signals reflecting off of the hand, and then analyzes the incoming RF signals to identify the micro-gesture.

Using rasterization to identify traffic signal devices

Systems and methods are provided for identifying and representing a traffic signal device. The method includes determining a location and orientation of the vehicle and receiving a real world image. The method further includes analyzing information about the vehicle's location and environment and using this information and the vehicle's orientation to generate a raster image illustrating an approximation of a view of the real world image, including one or more traffic signal devices. Additionally, the method includes providing the real world image and the raster image as inputs to a neural network to classify a traffic signal device in the real world image as the primary traffic signal device and determine a set of coordinates indicating a location of the primary traffic signal device, generating a classified real world image which includes a bounding box indicating the set of coordinates, and receiving the classified real world image.

Preprocessor training for optical character recognition

A method includes executing a Optical Character Recognition (OCR) preprocessor on training images to obtain OCR preprocessor output, executing an OCR engine on the OCR preprocessor output to obtain OCR engine output, and executing an approximator on the OCR preprocessor output to obtain approximator output. The method further includes iteratively adjusting the approximator to simulate the OCR engine using the OCR engine output and the approximator output, and generating OCR preprocessor losses using the approximator output and target labels. The method further includes iteratively adjusting the OCR preprocessor using the OCR preprocessor losses to obtain a customized OCR preprocessor.

Systems and methods of data augmentation for pre-trained embeddings

Systems and methods are provided for generating textual embeddings by tokenizing text data and generating vectors to be provided to a transformer system, where the textual embeddings are vector representations of semantic meanings of text that is part of the text data. The vectors may be averaged for every token of the generated textual embeddings and concatenating average output activations of two layers of the transformer system. Image embeddings may be generated with a convolutional neural network (CNN) from image data, wherein the image embeddings are vector representations of the images that are part of the image data. The textual embeddings and image embeddings may be combined to form combined embeddings to be provided to the transformer system.

Method and system for generating a fusion image

A method and system for generating a fusion image. The method may include obtaining a first image and a second image of a same scene. Pixels of at least a part of the first image may correspond to pixels of at least a part of the second image. The method may include generating, based on the first image, a first low-frequency image corresponding to a low-frequency component of the first image, the first low-frequency image including a plurality of first pixels. The method may include generating, based on the second image, a first high-frequency image corresponding to a high-frequency component of the second image, the first high-frequency image including a plurality of second pixels, the plurality of second pixels corresponding to the plurality of first pixels. The method may include determining, based on the first low-frequency image and the first high-frequency image, a featuring region in the first high-frequency image, wherein the featuring region and a region of the first image that ...

Semantic annotation of sensor data using unreliable map annotation inputs

A method of training a neural network, comprising: obtaining an image of a geographic area 502a; obtaining data representing unvalidated annotations for the geographic area 502b; obtaining data representing validated annotations for the geographic area 508; and training a neural network using the image and unvalidated data as inputs and the validated data as ground truth data 506, 510. The image of the geographic area may comprise a birds-eye view image, a ground-level image or a point cloud image. The unvalidated annotations for the geographic area may comprise crowd sourced (internet) annotations in the form of a graph representing drivable surfaces and nodes representing traffic intersections. The graph may be converted into a raster images and processed to fit the image data before the sensor data and unvalidated annotations are combined into a multi-layered image to be input to the neural network 504. The trained neural network may recognise: drivable surfaces; intersections; crosswalks ...

Step-by-step decomposition model-based charging station utilization rate prediction method and system

The invention relates to a charging station utilization rate prediction method based on a step-by-step decomposition model, and the method comprises the steps: collecting historical charging station utilization rate data, and corresponding weather data and date data as original data; constructing and training a step-by-step decomposition prediction model, and inputting the original data before the to-be-tested time period into the step-by-step decomposition prediction model to obtain a charging station utilization rate prediction value; the step-by-step decomposition prediction model comprises a deep information fusion layer and a residual decomposition prediction layer; the deep information fusion layer performs deep extraction and fusion on the input data by using an attention coding method to obtain an information fusion sequence and inputs the information fusion sequence to the residual decomposition prediction layer; the residual decomposition prediction layer decomposes the input ...

Drainage pipeline siltation disease dynamic diagnosis method based on SSAE and MLSTM

The invention discloses a drainage pipeline siltation disease dynamic diagnosis method based on SSAE and MLSTM, and the method comprises the steps: carrying out the sample collection of the siltation degree and siltation length index of a pipeline, and the flow velocity and flow change index data of an inlet and outlet through employing fluent, and carrying out the data preprocessing; constructing a pipeline siltation detection model based on a stack type sparse auto-encoder SSAE and an SVM classifier, carrying out feature extraction and siltation detection on water flow related parameters of the underground drainage pipeline, and outputting a siltation moment; constructing a multivariate long and short memory neural network MLSTM, carrying out siltation disease condition identification on the pipeline siltation moment data, and outputting a siltation condition category; according to the method, the dynamic defect of the underground drainage pipeline is diagnosed quickly and effectively ...

Asphalt road moving load identification method and device, electronic equipment and medium

The invention discloses an asphalt road moving load identification method and device, electronic equipment and a medium. The method comprises the steps that the load speed, the wheelbase and the asphalt layer temperature are acquired; correcting the central wavelength data by adopting a preset strain correction algorithm to obtain an asphalt road strain parameter; according to the vehicle speed and the asphalt layer temperature, a preset numerical simulation method is adopted, and the target load size and the target load transverse offset are determined; constructing an initial BP neural network model, and inputting the load speed, the wheelbase, the asphalt layer temperature and the strain parameters into the initial BP neural network model to obtain a target BP neural network model; and inputting the load speed, the wheelbase, the asphalt layer temperature, the strain parameter and the target load transverse offset into the target BP neural network model to obtain an identification result ...

Marine engine room intelligent operation and maintenance system based on human factor engineering

The invention discloses a marine engine room intelligent operation and maintenance system based on human factor engineering. The marine engine room intelligent operation and maintenance system is composed of a shipborne system, a shore-based system and a communication system. The shore-based system comprises a shore-based maintenance support center, a dispatching maintenance resource and a human factor risk perception system, and the shipborne system comprises cabin core equipment, a cabin equipment full life cycle state perception and evaluation system and a cabin resource management and control system. After state monitoring, information management and data analysis of ship system equipment are performed and various information is displayed and stored, the remote control capability of remote data storage, remote test, remote operation and maintenance and remote diagnosis is achieved at the same time, and real-time scheduling decision support is provided for cabin equipment operation and ...

Integrated learning electricity price prediction method and device based on distributed data fusion

The invention relates to an integrated learning electricity price prediction method based on distributed data fusion, and the method comprises the steps: collecting historical electricity price data of each city distributed in a provincial company region; performing data preprocessing on the historical electricity price data to obtain processed data; training the processed historical electricity price data through a machine learning model, and then performing electricity price prediction to obtain an initial prediction result of the electricity price; verifying the model after machine learning by using the processed historical electricity price data to obtain the total error of each city, and performing weighted fusion on the preliminary prediction result to obtain the future predicted electricity price of each city; and each city uploads the respective future predicted electricity price to the center for decision level fusion to obtain the future predicted total electricity price of the ...

Active and passive microwave soil moisture fusion algorithm based on generalized linear regression model

The invention provides an active and passive microwave soil moisture fusion algorithm based on a generalized linear regression model, and relates to the field of microwave remote sensing, and the method comprises the steps: obtaining current active microwave soil moisture data and current passive microwave soil moisture data; inputting the current active microwave soil moisture data and the current passive microwave soil moisture data to a soil moisture data fusion model, and obtaining active and passive microwave soil moisture fusion data output by the soil moisture data fusion model, the soil moisture data fusion model is determined according to the active microwave fusion weight and the passive microwave fusion weight; the active microwave fusion weight and the passive microwave fusion weight are determined according to a back scattering coefficient standardized regression coefficient and a brightness temperature standardized regression coefficient in a generalized linear regression ...

Construction method of literature and blog knowledge graph system

The invention relates to the technical field of literature and blog application, in particular to a literature and blog knowledge graph system construction method which comprises the steps of data collection, data processing, knowledge graph construction, knowledge graph application and cultural building knowledge graph system establishment during construction. According to the method and the system, excessive experts and human resources do not need to be consumed, the immersive experience of a user is improved while three-dimensional atlas visualization, virtual human intelligent interaction and knowledge acquisition of the user are realized, the 3D atlas and literature and blog knowledge are combined, multiple abundant atlas analysis applications including knowledge retrieval, intelligent question and answer, knowledge personalized recommendation and information aggregation are integrated, and the user experience is improved. And the ability of knowledge display and knowledge spreading ...

Troposphere height difference correction model construction method based on multi-source data fusion

The invention relates to a troposphere height difference correction model construction method based on multi-source data fusion. When the height difference between observation stations is large, troposphere delay changes are complex, and an existing troposphere model is low in correction precision. According to the method, time and space reference conversion is carried out on data of GNSS, COSMIC, radio sounding station, ECMWF and ITG models; constructing a background field by using the data of the ECMWF, and detecting and eliminating gross error anomalies; calibrating the systematic deviation by using the data of the GNSS; determining multi-source data weights under different elevations based on a quadratic unbiased estimation method by taking the change trend of the medium and long term troposphere delay signal as a constraint; and calculating a height reduction function of each grid point by using bilinear interpolation, and establishing a troposphere height difference correction model ...

Active and passive radar target information fusion method and device

The invention discloses an active and passive radar target information fusion method and device. The method comprises the following steps: respectively obtaining target measurement information and target attribute information detected by an active radar and a passive radar; associating targets detected by the active radar and the passive radar according to the target measurement information to obtain an associated observation pair; fusing each target in each correlation observation pair by using an evidence theory method, taking the angle correlation confidence coefficient and the category confidence coefficient as two evidences when the evidence theory method is used for fusion, and obtaining the fusion confidence coefficient output of each target after fusion so as to realize the active radar and passive radar target fusion. The angle correlation confidence coefficient is obtained by transforming the statistical magnitude of the angle deviation between the targets detected by the active ...

Eye using habit scoring method and system

The invention discloses a scoring method and system about eye using habits. Unified quantitative scoring is carried out on the eye using habits from the two aspects of eye using behaviors and eye using environments. For eye using behaviors, collecting eye using behavior entry data such as visual objects, visual distances and visual duration, and then performing weighted statistics on eye using behavior entries through a scale matrix to obtain an eye using behavior score; according to the eye use environment, safe light environment information is calculated according to the environment light intensity change curve and the light warning curve corresponding to the light intensity change curve, and therefore an eye use environment score is obtained according to the probability distribution density; and finally, performing weighted calculation on the eye using behavior score and the eye using environment score to obtain a comprehensive quantitative analysis result of the eye using habits. The ...

Multi-source data fusion method and system, electronic equipment and storage medium

The invention relates to a multi-source data fusion method and system, electronic equipment and a storage medium. The fusion method comprises the following steps: aligning measurement data to be fused with a timestamp of an existing effective fusion target; performing weighted summation according to the measurement data and the mahalanobis distance of the position and the Euclidean distance of the speed of the aligned effective fusion target to obtain a weighted distance, and calculating an association threshold range and an association probability of the measurement data and the aligned effective fusion target based on the weighted distance; according to the association threshold range and the association probability, judging whether the measurement data is associated and matched with any effective fusion target to generate a matching pair; in each matching pair, the filtering algorithm in the effective fusion target is updated by using the measurement data; the Mahalanobis distance of ...

Information matching method, device and equipment and readable storage medium

The invention discloses an information matching method, device and equipment and a readable storage medium, relates to the technical field of communication, and aims to improve the matching degree between customer service staff and customers. The method comprises the following steps: acquiring information of a target customer service and information of candidate customers; taking the information of the target customer service and the information of the candidate customers as input of a matching model, operating the matching model, and determining a target customer from the candidate customers; outputting the information of the target customer; wherein the matching model is obtained according to the first fusion feature vector of the reference customer service and the second fusion feature vector of the reference customer. According to the embodiment of the invention, the matching degree between the customer service and the customer can be improved.

Car-following model considering information fusion of rear car and multiple front cars and based on intelligent network connection

The invention relates to the technical field of vehicle track prediction, and discloses an intelligent network connection-based car-following model considering information fusion of a rear vehicle and multiple front vehicles, and the expression is as follows: on the basis of an OVM model and an FVD model, a car-following model is obtained; the coupling effect of factors such as a vehicle sensor sensitivity coefficient, front and rear vehicle speed collaborative optimization, a multi-front vehicle speed difference value and a multi-front vehicle headstock distance is comprehensively considered, a traffic flow vehicle following improved model for multi-vehicle information fusion in an intelligent network connection environment is constructed, critical stability conditions are deduced, parameter sensitivity numerical simulation is carried out, and the accuracy of the traffic flow vehicle following improved model is improved. The fluctuation influence effect of each index coefficient on the ...

Multi-sensory intelligent accompanying robot system

The invention relates to a multi-sensory intelligent accompanying and accompanying robot system which comprises a multi-source heterogeneous fusion intelligent framework, a monitoring system model and an intelligent accompanying and accompanying robot. The intelligent accompanying robot comprises an FPGA core processing module, a computer image processing module, a CMOS driving circuit, a motion tracking module, a power supply module, an automatic monitoring charging module, a storage module, a debugging module, a network transmission module, a multi-sensory sensor and an unmanned aerial vehicle module. The multi-point interaction technology of the unmanned aerial vehicle and the intelligent robot is adopted, the unmanned aerial vehicle achieves real dead-corner-free monitoring through a high-altitude view angle, meanwhile, multi-dimensional sample data can be captured, data support is provided for the robot system to make a correct decision, the accuracy and reliability of the system are ...

Semi-supervised protein lysine post-translational modification site prediction method and system

The invention provides a semi-supervised protein lysine post-translational modification site prediction method and system, and the method comprises the steps: carrying out data collection and preprocessing, and constructing a data set; designing a network structure, and constructing a multi-scale sequence feature extraction network; training the model by using a label-free sample self-distillation method; and the prediction result is evaluated by adopting indexes such as accuracy and recall rate. The technical problems that no-label data cannot be effectively utilized, knowledge migration between species is difficult, and the robustness of a species specificity prediction model is low are solved.

Rolling bearing fault diagnosis method based on parallel convolutional neural network

The invention discloses a rolling bearing fault diagnosis method based on a parallel convolutional neural network. The rolling bearing fault diagnosis method comprises the following steps: S100, acquiring an original vibration signal Xio (t) of a rolling bearing to be detected; s200, preprocessing the original vibration signal Xio (t) to obtain a preprocessed vibration signal Xi (t); s300, performing first overlapping sampling on the preprocessed vibration signal Xi (t) to obtain a first sample signal xm (t); s400, performing second overlapping sampling on the preprocessed vibration signal to obtain a second sample signal ym (t); s500, respectively obtaining a one-dimensional tensor set and a two-dimensional tensor set of the rolling bearing to be tested based on the first sample signal xm (t) and the second sample signal ym (t); and S600, constructing a fusion data set based on the one-dimensional tensor set and the two-dimensional tensor set of the rolling bearing to be detected, and ...

Meteorological disaster high-resolution regional mode forecasting system and method

The invention provides a meteorological disaster high-resolution regional mode forecasting system and method, and relates to the technical field of meteorological disaster forecasting, and the method comprises the steps: carrying out the downscaling of target data of a to-be-processed region, obtaining the downscaled target data, carrying out the fusion of multi-source observation data, and obtaining the grid point fusion data; performing physical quantity correction, error estimation, quality control and format conversion processing on different target observation data by adopting a specific technical method according to a preset period, and performing hour-by-hour rapid cycle assimilation processing on the processed target observation data to obtain target assimilation analysis data; using the downscaled target data, the target assimilation analysis data and the localized data to simulate and operate a localized numerical forecasting mode, generating meteorological disaster numerical ...

Bridge deflection monitoring method based on Beidou navigation, electronic equipment and storage medium

The embodiment of the invention provides a bridge deflection monitoring method based on Beidou navigation, electronic equipment and a storage medium. The method comprises the following steps: processing second bridge deformation data of an accelerometer to obtain high-frequency deformation data and a low-frequency feature vector; processing the first bridge deformation data of the Beidou to obtain low-frequency deformation data and a high-frequency feature vector; calculating a first correction parameter and a second correction parameter; and finally, according to the first correction parameter, the second correction parameter, the low-frequency deformation data and the high-frequency deformation data, determining bridge deflection information of the target bridge. High and low frequency analysis is carried out on monitoring results of a Beidou system and an accelerometer, a correction coefficient is calculated according to the analysis result to represent the monitoring difference of the ...

Aviation gamma energy spectrum data fusion processing method and system

The invention discloses an aviation gamma energy spectrum data fusion processing method and system. According to the method, each energy window count of a gamma energy spectrum is composed of a full energy peak count and a Compton scattering background count; the range of each energy window of a CeBr3 detector with high energy resolution is taken as a datum reference, counting rate data of each energy window acquired by a large-volume NaI (T1) detector with high detection efficiency is corrected into a Compton scattering background with the same narrow energy range as that of each energy window of the CeBr3 detector, and the counting rate data of each energy window of the two types of detectors are combined; and processing the merged energy window data by adopting a unified correction coefficient to obtain a final natural nuclide content value. The method has the advantages that the dual advantages of high energy resolution of the two detectors are exerted, the dual targets of high-efficiency ...

Aquaculture water dissolved oxygen data restoration method based on data fusion

The invention discloses an aquaculture water dissolved oxygen data restoration method based on data fusion, and relates to the technical field of aquaculture, the implementation of the method specifically comprises the following steps: data acquisition: in a deployed aquaculture wireless sensor network environment, performing data acquisition; index data such as dissolved oxygen, water temperature and pH value of seven monitoring nodes in the pond are collected according to actual production requirements; data detection and sample preparation: performing consistency detection on the collected multi-sensor data to obtain data after smooth processing, and preparing a multi-monitoring-point isomorphic sensor sample set; self-support degree calculation: calculating the dissimilarity degree and correlation degree among all terminal sensor nodes by using dissolved oxygen smooth data of multiple sensors in the aquaculture water, and measuring the self-support degree value between every two sensors ...

Multi-channel information fusion piano music notation method based on cyclic notation

The invention relates to a multichannel information fusion piano music score recording method based on a cyclic spectrum, which comprises the following steps of: inputting the cyclic spectrum of a piano signal into a detection network with different targets, and respectively detecting a starting point, an ending point and a note duration of a piano note; fusing three channel features of a piano note starting point, a note ending point and a note continuous section to obtain an enhanced feature, and performing dot product operation on the enhanced feature matrix and three feature matrixes of the note starting point, the note ending point and the note continuous section to obtain a fused feature matrix; extracting cross-channel information of the fusion feature matrix by using a convolutional neural network; the time sequence relevance of cross-channel information is modeled by using a bidirectional gating loop unit, and a piano music note sequence is obtained by a pitch sequence probability ...

Equipment state control method and device, storage medium and electronic device

The invention discloses an equipment state control method and device, a storage medium and an electronic device. The equipment state control method comprises the following steps: acquiring current area sensing data collected by a group of sensing sensors in a target area at the current moment; analyzing the current object behavior characteristics and the current area environment characteristics from the current area perception data to obtain area perception characteristics at the current moment, and fusing the object behavior characteristics and the area environment characteristics at the same moment to obtain area perception characteristics; according to the region perception feature sequence, predicting the region perception feature at the next moment, the region perception feature sequence comprising a group of historical region perception features and the current region perception feature, and the predicted region environment feature in the predicted region perception feature representing ...

Data management system based on data fusion

The invention relates to the technical field of data management, in particular to a data management system based on data fusion, which comprises a data acquisition module, a data extraction module, a data processing module, a data fusion module and a data recommendation module which are connected according to a linear sequence, the data recommendation module generates standard user features and standard product features, the data fusion module performs feature fusion on the standard user features and the standard product features to obtain standard recommendation features and generates a standard recommendation feature group according to all the standard recommendation features, and the data recommendation module calculates product recommendation values of products according to the standard recommendation feature group. According to the method and the system, the product recommendation value is obtained, the product with the product recommendation value larger than the recommendation threshold ...

Automobile hub bearing diagnosis method based on adaptive cross fusion adversarial learning network

The invention discloses an automobile hub bearing diagnosis method based on an adaptive cross fusion adversarial learning network. The method comprises the following steps: step 1, collecting vibration signals and pulse signals of six types of automobile hub bearings in a fault failure scene by using a vibration sensor and a tachometer, and obtaining model training data by using a CSCOT-FFT preprocessing algorithm; step 2, pre-training GSAE by using the data obtained in the step 1 to obtain a cross information fusion layer; step 3, building an ACFALN model in combination with the cross information fusion layer obtained by training in the step 2, and training until the model converges; and 4, applying the trained ACFALN model to reality, carrying out real-time state identification on an actual sampling signal, and displaying an alarm on data with abnormal states on a vehicle-machine interface to ensure the safety of people and vehicles. According to the method, the abnormal state of the ...

Radar frequency band fusion extension method based on MP and Root-MUSIC algorithm

The invention belongs to the technical field of radar target identification, and particularly relates to a radar frequency band fusion extension method based on MP and Root-MUSIC algorithms. Comprising the following steps of: 1, setting two adjacent radars with the same observation angle and different working frequencies, so as to establish an all-pole GTD model of low-frequency and high-frequency electromagnetic echoes for a stationary target consisting of scattering centers; step 2, incoherent parameters, namely a linear phase and a fixed phase, are estimated by using the echo data and conjugate data thereof; and step 3, processing the coherent data by using an improved algorithm, estimating a pole and an amplitude of the broadband complete polarization model, and then calculating by using the original data of the low-frequency sub-band and the high-frequency sub-band to obtain a full-band fusion signal. According to the multiband fusion method provided by the invention, the new covariance ...

Equipment health assessment method and system based on group data and auto-encoder

The invention provides an equipment health assessment method and system based on group data and an auto-encoder. The method comprises the following steps: acquiring operation state signals of a plurality of same equipment under the same working condition; preprocessing all the signals, calculating differences among the signals and generating difference matrixes, and splicing the difference matrixes with multiple moments and multiple time spans into a tensor; using a tensor spliced by the difference matrix as training data, training a multi-scale auto-encoder to minimize a reconstruction error, and generating a machine learning model; and inputting the processed equipment operation data into a multi-scale auto-encoder, reflecting the degree of deviation of the data from a normal state through a reconstruction error, and evaluating the health condition of the equipment according to the abnormal degree of the data. Abnormity of individual data is judged according to the difference between ...

Resource recommendation method and device and storage medium

The invention provides a resource recommendation method and device and a storage medium, and relates to the technical field of computers.The method comprises the steps that vector clustering processing is carried out on word vectors, aggregation processing is carried out on word vector groups corresponding to resources, resource feature vectors are generated in a semantic clustering and vector fusion mode, and the resource feature vectors are stored in a database; a knowledge graph is constructed, propagation preferences of all resource feature vectors in the knowledge graph are obtained through attention mechanism training, training learning is carried out based on interaction relations between all kinds of resources and user types, and multiple prediction models are trained; and performing resource recommendation processing by using the knowledge graph and/or the trained prediction model. The information of the resource entity and the interaction information can be fully utilized for ...

Patent text entity extraction method and device, equipment and medium

The invention discloses a patent text entity extraction method and device, equipment and a medium, and the method comprises the steps: generating at least one triple according to a claim subject, a claim reference relationship and a claim serial number in a claim text, and generating a reference relationship topological graph corresponding to the claim subject according to the triple; inputting the claim text corresponding to each reference relation topological graph into a pre-training model bert to generate a first entity pair; dividing the claim text corresponding to each reference relation topological graph into at least one technical scheme block according to the reference relation, and respectively inputting the technical scheme blocks into a pre-training model bert to generate corresponding technical scheme entity pairs; splicing the technical scheme entity pairs to generate a second entity pair; and cutting and fusing the first entity pair and the second entity pair to generate ...

Intelligent foundation pit monitoring method based on multi-sensor data fusion

The invention discloses a foundation pit intelligent monitoring method based on multi-sensor data fusion, and belongs to the technical field of foundation pit intelligent monitoring. According to the invention, the estimation information of a local filter LF and a main filter MF is fused into new global state estimation information by constructing a multi-sensor data fusion time-varying model, sensors of each foundation pit monitoring subsystem and filters of a main sensor, carrying out independent time updating on each filter and carrying out measurement updating on a local filter; and obtaining the latest monitoring state value of the foundation pit. According to the method, data of various sensors used in foundation pit monitoring are comprehensively applied, theoretical values, measured values and change information of all foundation pit monitoring sensors in the front and back monitoring periods are fused, complementation of the foundation pit monitoring data is achieved, abnormal ...

Environmental water quality monitoring system and monitoring method thereof

The invention relates to the field of water quality safety, and particularly discloses an environmental water quality monitoring system and a monitoring method thereof, which are characterized in that global-based high-dimensional semantic features among various water quality detection data are extracted through a context encoder so as to be more suitable for characterizing essential mode features of the water quality; and extracting a dynamic change implicit rule of the water quality by using a convolutional neural network model and a time sequence encoder so as to perform feature fusion through a Gaussian density map and a Gaussian mixture model, and considering that feature distribution of features in a feature space has position sensitivity relative to a label value in a classification process so as to obtain a classification result. Therefore, label value scattering response factors of the first feature matrix and the second feature matrix are respectively calculated as weighting coefficients ...

Image feature combination for image-based object recognition

Methods, systems, and articles of manufacture to improve image recognition searching are disclosed. In some embodiments, a first document image of a known object is used to generate one or more other document images of the same object by applying one or more techniques for synthetically generating images. The synthetically generated images correspond to different variations in conditions under which a potential query image might be captured. Extracted features from an initial image of a known object and features extracted from the one or more synthetically generated images are stored, along with their locations, as part of a common model of the known object. In other embodiments, image recognition search effectiveness is improved by transforming the location of features of multiple images of a same known object into a common coordinate system. This can enhance the accuracy of certain aspects of existing image search/recognition techniques including, for example, geometric verification.

Information processing apparatus, information processing method, and program

An information processing apparatus, an information processing method, and a program, the information processing apparatus and the information processing method appropriately combining observation values of a plurality of sensors according to noise characteristics of the plurality of sensors and a condition for the observation values. An Allan variance is calculated as the characteristics of each of the plurality of sensors. A time window length for an Allan variance is specified as a condition used when the observation values of the plurality of sensors are combined. A weight for each of the plurality of sensors is calculated such that a sum of the Allan variances for the specified time window length is smallest. The respective observation values of the plurality of sensors are combined using a sum of products of the observation values of the plurality of sensors and the weights calculated for each corresponding sensor of the plurality of sensors.

ATTENTIONAL SAMPLING FOR LONG RANGE DETECTION IN AUTONOMOUS VEHICLES

Methods and systems for detecting objects near an autonomous vehicle (AV) are disclosed. An AV will capture an image. A trained network will process the image at a lower resolution and generate a first feature map that classifies object(s) within the image. The system will crop the image and use the network to process the cropped section at a higher resolution to generate a second feature map that classifies object(s) that appear within the cropped section. The system will crop the first feature map to match a corresponding region of the cropped section of the image. The system will fuse the cropped first and second feature maps to generate a third feature map. The system may output the object classifications in the third feature map to an AV system, such as a motion planning system that will use the object classficiations to plan a trajectory for the AV.

Three-dimensional coordinate fusion positioning method and system and computer equipment

The invention relates to the technical field of coordinate positioning, and provides a three-dimensional coordinate fusion positioning method and system and computer equipment, and the method comprises the steps: collecting the space coordinates of a target user in a target scene in a three-dimensional visual model; wherein the three-dimensional visual model is a visual model constructed for the target scene; collecting image information of a target user in the target scene, and determining space plane coordinates of the target user from the image information; and fusing the space coordinate of the target user in the three-dimensional visual model, the space plane coordinate of the target user and the timestamp of the image information to obtain a fused positioning coordinate of the target user in the three-dimensional visual model. According to the method, the space coordinates of the target user are collected, the space plane coordinates in the image information and the timestamp of the ...

Data acquisition method, system and related device

The invention discloses a data acquisition method and system and a related device. The data acquisition method can be applied to a data acquisition system, the data acquisition system comprises a plurality of devices and a federation management center, and the method comprises the following steps: the federation management center can receive a data acquisition task sent by a first device in the plurality of devices; the federal management center determines one or more devices for providing sensing data required by the data acquisition task according to the data acquisition task and the sensing capability information of the plurality of devices; the federal management center instructs one or more devices to provide required sensing data. According to the data acquisition method provided by the embodiment of the invention, a plurality of (two or more) electronic devices can establish a data acquisition system, and one electronic device in the data acquisition system can acquire original data ...

Alzheimer's disease prediction method fusing fMRI image and genomics data

The invention relates to the technical field of genomics data, in particular to an Alzheimer's disease prediction method fusing fMRI images and genomics data, and provides the following scheme for solving the problems of long model training time, long multi-feature classification period, low multi-feature classification accuracy and the like in the prior art. Comprising the following steps: S1, preprocessing a functional image: performing AC correction and segmentation on an Fmri image obtained from ADNI, registering the Fmri image to a standard space, and calculating a roi signal; the invention aims to detect the relationship between the brain region and the gene by performing a correlation analysis method on different forms of high-dimensional data, samples in a selected data set have universality, a proposed feature construction method and a random forest-Apriori model have good applicability and classification capability, gene information and image data are fused, and the accuracy of ...

Data-model back-end fusion method for automatic driving decision control function

The invention discloses a data-model back-end fusion method for an automatic driving decision control function. The method comprises the following steps: step 1, acquiring driving data and constructing a data set; 2, establishing and verifying a vehicle dynamics theoretical model; 3, constructing an automatic driving decision control problem into an integrated dynamic optimization problem, and solving the decision control problem at each moment; 4, performing model prediction rolling time domain online optimization solution on the dynamic optimization problem; step 5, performing data-driven reinforcement learning off-line optimization solution on the dynamic optimization problem; 6, calculating the confidence coefficient of pure model driven solution; 7, calculating the confidence coefficient of pure data driven solution; and 8, calculating a rear-end fusion factor and a fusion decision control quantity of model data information. According to the method, the safety and the decision control ...

Fraud user identification method and device, electronic equipment and storage medium

The invention relates to the technical field of big data application, and provides a fraud user identification method and device, electronic equipment and a storage medium. The method comprises the following steps: acquiring a multi-dimensional feature index of a user; the multi-dimensional characteristic indexes comprise characteristic indexes related to a black base station; according to the multi-dimensional feature index of the user and a multi-classifier fusion recognition model, obtaining a recognition result that the user belongs to a fraud user; wherein the multi-classifier fusion recognition model is obtained by training a plurality of base models on the basis of multi-dimensional feature indexes of samples with determined classification labels. According to the fraud user identification method and device, the electronic equipment and the storage medium provided by the embodiment of the invention, the prediction accuracy of the model can be improved and the robustness of the model ...

Multi-mode autonomous underwater vehicle state detection method and system

The invention relates to a multi-mode autonomous underwater vehicle state detection method and system. Belongs to the technical field of state detection, and particularly relates to the technical field of autonomous underwater vehicle state detection. The problems that existing abnormal state detection cannot effectively cope with multi-modal sensor data, and the difference between different modal data is ignored are solved. The method comprises the following steps: S1, inputting sensor data of each system of the autonomous underwater vehicle; s2, weighting the signal of each mode through extrusion and excitation operation, and then scaling; s3, modal features are extracted through a self-attention mechanism with residual errors; s4, performing feature fusion on different modes through extrusion and excitation operation; and S5, performing decision classification on the fused features, and outputting the prediction probability of each category. The method and the system can be applied to ...

Cigarette formula tobacco leaf replacement method fusing near infrared spectrum thermal cracking components

The invention discloses a cigarette formula tobacco leaf replacement method fusing near-infrared spectrum thermal cracking components, which comprises the following steps of: (1) performing near-infrared characteristic spectrum section data and thermal cracking product characteristic component data of tobacco leaves in a cigarette formula by applying a multi-modal fusion algorithm based on a convolutional neural network; obtaining a tobacco leaf fusion feature matrix X in the cigarette formula; (2) applying a multi-modal fusion algorithm based on a convolutional neural network to near-infrared characteristic spectrum data and thermal cracking product characteristic component data of different stock tobacco leaves to obtain a stock tobacco leaf fusion characteristic matrix Y; (3) calculating a mahalanobis distance value d of the matrixes X and Y; and (4) selecting the stock tobacco leaves with the mahalanobis distance value d not greater than a certain value as the substitute tobacco leaves ...

Bio-medical text overlapping relation joint extraction method and related equipment

The invention provides an overlapping relation joint extraction method and related equipment for biomedical texts, and the method comprises the steps: constructing an overlapping relation triple joint extraction framework with a fusion module, a subject extraction module and an object extraction module through obtaining to-be-extracted biomedical text data, inputting to-be-extracted biomedical text data into the overlapping relation triple joint extraction framework for extraction to obtain an overlapping relation triple of the biomedical text data; compared with the prior art, each text in the biomedical text data to be extracted is fused with a plurality of relation types through the fusion module; the information of the starting position of the main body is introduced when the ending position of the main body is extracted to help to extract the main body; meanwhile, when the object is extracted, the information of the starting position of the object and the relative distance information ...

Industrial anomaly detection and solution knowledge graph construction method

The invention provides an industrial anomaly detection and solution knowledge graph construction method, which comprises the following steps: acquiring enterprise equipment data; entities and relationships among the entities are extracted from the enterprise equipment data, and an enterprise equipment knowledge graph is constructed; obtaining phenomenon data occurring in the production process of the enterprise products, and establishing a logical relationship network of the enterprise products and the enterprise equipment based on the enterprise knowledge equipment graph and the phenomenon data; realizing communication of an enterprise equipment knowledge graph and a logic relation network through fusion between nodes of a physical layer and nodes of a logic layer, and establishing a phenomenon-gt; a reason-gt; the data chain of the solution is used for complementing the enterprise equipment knowledge graph; and performing anomaly detection in industrial production based on the complemented ...

Pilot operation state monitoring method based on multilayer fusion recognition model

The invention provides a pilot operation state monitoring method based on a multilayer fusion recognition model. The pilot operation state monitoring method comprises the steps that S1, peripheral physiological feature data, wrist acceleration time sequence data and wrist acceleration feature set data based on different sensors are collected; s2, constructing an SVM recognition model based on an RBF kernel function; s3, establishing an SVM recognition model according to the optimal parameters; s4, constructing an operation load identification model based on CNN-LSTM and intermediate layer fusion; s5, training to obtain a CNN-LSTM operation load identification model; s6, constructing a dynamic weighted fusion mode of a decision-making layer; and S7, performing state recognition by using the trained recognition model. According to the method, the KPCA dimension reduction technology is adopted, the state information of the pilot in different dimensions can be fully perceived by inputting the ...

Classification model training method and related device

A classification model training method is applied to the technical field of artificial intelligence. In the method, target training data with smaller data volume is generated based on original training data to obtain training data of different scales, and the scale diversity of the training data is ensured; moreover, in consideration of the characteristics of the AI generated data, the target training data which is small in data size and originally generated by the AI is marked as label-free data, so that the classification model can focus on the data generated by the AI and different from the data generated manually during training, and the classification accuracy is improved. The influence on the perception of the classification model on the AI generation data style when the AI generation data is marked as the AI generation data is avoided, and the prediction accuracy of the classification model is effectively improved.

Construction method of crop planting model and crop planting method

The embodiment of the invention is suitable for the technical field of agriculture, and provides a construction method of a crop planting model and a crop planting method.The construction method of the crop planting model comprises the steps that various first-level index information related to crop planting is obtained, any first-level index information is provided with a corresponding index type, and the index type of the first-level index information corresponds to the index type of the first-level index information; and according to the plurality of primary index information and the index types, generating a plurality of secondary index information, respectively adopting the plurality of secondary index information to construct a plurality of secondary models, and fusing the plurality of secondary models to obtain a crop planting model. By applying the crop planting model constructed according to the method, scientific guidance and timely early warning can be provided for planters in ...

Speech training feedback method and device and storage medium

The invention discloses a speech training feedback method and device and a storage medium, image data and sound data of a speaker are acquired in real time through an acquisition device, the image data comprises three-dimensional depth information of a space where the speaker is located, subsequent processing analysis is performed based on multi-modal data, and more accurate data support can be provided; and performing first processing on the image data in real time based on an adaptive enhancement model and a regional attention mechanism through the edge device, performing second processing on the sound data based on an adaptive suppression model and a voice activity detection model, and performing multi-modal fusion recognition on a first processing result and a second processing result. Higher processing speed is realized, and the real-time performance is ensured; and the target speech effect evaluation model is matched to output the improvement suggestion, so that the real-time performance ...

Data fusion algorithm of multi-police joint service control system

The invention discloses a data fusion algorithm of a multi-police joint service control system, and relates to the technical field of communication processing, and the algorithm specifically comprises the following steps: S1, collecting and extracting alarm position data and nearby rescue data; s2, according to the alarm position data, substituting the alarm position data into a threat value calculation strategy to calculate a threat value of an alarm position, and according to the nearby rescue data, substituting the nearby rescue data into a rescue value calculation strategy to calculate a rescue value of each nearby rescue unit; s3, the threat value of the alarm position and the rescue value of each nearby rescue are substituted into a rescue scheme generation strategy, and a rescue scheme is generated; and S4, rescue tasks are allocated according to the generated rescue scheme, and rescue information is issued to the corresponding rescue units, so that the rescue units nearby can be ...

Power distribution network main equipment fault probability model parameter identification method and system

The invention discloses a power distribution network main equipment fault probability evaluation model parameter identification method and system, and the method comprises the steps: building a fault probability evaluation model, obtaining a maximum fault probability and a minimum fault probability from fault probability historical data, and enabling the maximum fault probability to correspond to a health state first evaluation value, performing first identification on the to-be-identified parameter by using the minimum fault probability corresponding to the second evaluation value of the health state; constructing an objective function by using the fault probability historical data and the health state historical data of the main equipment, and performing secondary identification on the to-be-identified parameters through iterative calculation by adopting optimization algorithms such as a genetic algorithm and taking the minimum value of the objective function as an optimization target ...

Method and device for monitoring surface subsidence of mining area and electronic equipment

The invention provides a mining area surface subsidence monitoring method and device and electronic equipment. The method comprises the steps of obtaining multiple pieces of position data of the earth surface of a mining area; performing first data fusion on the multiple pieces of position data to obtain an initial data set; according to the position data corresponding to different areas in the mining area, performing second data fusion on the data of the corresponding areas in the initial data set to obtain a target data set; and detecting the subsidence degree of the earth surface of the mining area according to the target data set. According to the scheme, the multiple pieces of position data of the earth surface of the mining area can be obtained, the target data set can be obtained according to the mapping relation among the multiple pieces of position data in a two-time data fusion mode, the target data set can well adopt the relevance among different pieces of position data, and ...

ELECTRONIC DEVICE EMPLOYING THERMAL SENSOR AND IMAGE SENSOR

There is provided a recognition system adaptable to a portable device or a wearable device. The recognition system senses a body heat using a thermal sensor, and performs functions such as the living body recognition, image denoising and body temperature prompting according to detected results.

System and method for monitoring touch-screen gestures of users and for user authentication

The present invention relates to an improved method of providing identification of a user or authentication of a user's identity. More particularly, the present invention relates to an improved method of providing identification of a user or authentication of a user's identity using conditional behavioural biometrics. The present invention seeks to provide an enhanced method of authenticating and/or identifying a user identity using conditional behavioural biometrics. According to a first aspect of the present invention, there is provided a method of generating a user profile for use in identifying and/or authenticating a user on a device, the device equipped with one or more sensors, the method comprising: generating a set of data points from sensory data collected by the one or more sensors; clustering the set of data points to produce a set of data clusters; developing a first classifier for the data clusters, the first classifier being operable to assign a further data point derived ...

Information processing device, information processing method and computer readable storage medium

An information processing device, an information processing method, a computer readable storage medium are provided. The information processing device comprises processing circuitry configured to: construct, for each of a plurality of indexes, a sample unit set for the index based on a plurality of minimum labeled sample units related to the index which are obtained and labeled from an original sample set; and extract, for at least a part of the constructed plurality of sample unit sets, a minimum labeled sample unit from each sample unit set, and generate a labeled training sample based on the extracted minimum labeled sample unit. A sample unit set is constructed based on minimum labeled sample units that are labeled manually, and a labeled training sample is generated automatically based on such sample unit sets, thereby generating the labeled training sample automatically to a certain degree, and reducing manual participation.

COMPUTER-IMPLEMENTED METHOD FOR 3D LOCALIZATION OF AN OBJECT BASED ON IMAGE DATA AND DEPTH DATA

Traffic accident prediction method

The invention discloses a traffic accident prediction method. The method specifically comprises a traffic accident prediction method based on system identification, a traffic accident prediction method based on LSTM, a traffic accident prediction method based on system identification and LSTM residual error combination, a traffic accident prediction method based on system identification and LSTM linear weighting combination, and a traffic accident prediction method based on system identification and LSTM streamline. According to the method, a traffic accident prediction method based on system identification or a traffic accident prediction method based on LSTM is combined or the two prediction methods are combined in different modes, prediction comparison is carried out, and results show that the different methods can improve the prediction capability of road traffic accidents; compared with an existing method, the method has higher prediction precision.

Virtual-real fusion traffic simulation experiment platform and method based on digital twinning

The invention relates to a virtual-real fusion traffic simulation experiment platform and method based on digital twinning. The simulation experiment platform comprises traffic simulation software, a control console, a data center, a digital sand table and a control console. The data center is respectively connected with the traffic simulation software and console and the digital sand table and console through a Client/Server structure, the data center comprises a server, and the traffic simulation software and console and the digital sand table and console comprise clients; the Client/Server structure is used for transmitting an information stream including vehicle state information and infrastructure information. Compared with the prior art, the virtual-real fusion mapping interaction between the traffic simulation software and the digital traffic sand table is realized through the digital twinborn technology, the traffic scene simulation enhancement can be realized, and the credibility ...

Multi-target track fusion method based on multiple sensors

The invention discloses a multi-target track fusion method based on multiple sensors. The method comprises the steps that the number n of tracking targets, namely to-be-fused tracks, is determined; classifying the data acquired by the m sensors by using a KMeans clustering algorithm to obtain n groups of data; performing Kalman filtering on the n groups of classification data of each sensor platform; with the earliest sampling starting time as the starting time and the latest sampling ending time as the ending time, performing interpolation on each sensor platform with the same timestamp to obtain measurement data with the same data size; respectively fusing the n groups of target tracks of the m sensors by using a convex combination fusion algorithm; when new data is measured, judging which track the current data belongs to based on a KMeans clustering algorithm; new observation points are classified to corresponding tracks, and then Kalman filtering is carried out again to complete iterative ...

Text content and image analysis combined multi-modal content retrieval method and system

The invention discloses a multi-modal content retrieval method combining text content with image analysis. Comprising the following steps: preprocessing a data set to obtain a text image information pair; extracting image and text features, and performing multi-modal attention calculation on the image features and the text features to obtain multi-modal features; encoding the image, the text and the multi-modal features to form corresponding hash codes; constructing a target loss function, and training to obtain a multi-modal hash code generation model; constructing a multi-modal hash code database of the database from the database to be retrieved by utilizing a multi-modal hash code generation model; and according to text information input by a user, generating a multi-modal hash code, and matching the multi-modal hash code with the multi-modal hash code database to obtain a retrieval result. The invention further discloses a multi-mode content retrieval system combining the text content ...

Medical image feature mining method for central nervous tumor and related device

The invention discloses a medical image feature mining method for central nervous tumors. The method comprises the following steps: acquiring an image map of preoperative magnetic resonance imaging scanning and a corresponding diagnosis text; performing three-dimensional image feature extraction operation on the image map to obtain a first feature set; performing radiomics feature extraction operation on the image to obtain a second feature set; performing semantic feature extraction operation on the diagnosis text to obtain a third feature set; and comprehensively processing the first feature set, the second feature set and the third feature set and then outputting a mining result. Through combination of radiomics feature extraction and semantic feature extraction of a diagnosis text, errors generated by image feature extraction completely based on machine learning are overcome to a certain extent, the marking accuracy of image features can be improved in combination with a medical background ...

Monitoring target fusion method and device based on multiple sensors, and electronic equipment

The invention discloses a monitoring target fusion method and device based on multiple sensors and electronic equipment, and the fusion method comprises the steps: carrying out the particle sampling of a state estimation value of a target state of each monitoring target at a previous moment and a contour estimation value of a target contour, the method comprises the steps of obtaining weighted particles of predicted measurement values of monitoring targets at the current moment, then processing measurement data of each sensor to obtain current state estimation values and current contour estimation values of all the monitoring targets corresponding to the sensors at the current moment, and then screening a target sensor set corresponding to each monitoring target, and fusing the current state estimation value and the current contour estimation value corresponding to each target sensor in the target sensor set to obtain a target state value and a target contour value of the monitoring target ...

Target object identification method, and training method and device of multi-modal identification model

The invention discloses a target object recognition method, a multi-modal recognition model training method and a related device, and relates to artificial intelligence, a computer vision technology, natural language processing and machine learning. After to-be-recognized image data and to-be-recognized text data of a to-be-recognized object are obtained, an initial recognition result of the to-be-recognized object is determined according to text features obtained by performing feature extraction on the to-be-recognized text data through the text sub-model or image features obtained by performing feature extraction on the to-be-recognized image data through the image sub-model; if the confidence coefficient of the initial recognition result is smaller than or equal to a first preset confidence coefficient or is not a preset result, a prediction recognition result of the to-be-recognized object is determined through a fusion result obtained by performing feature fusion on the text features ...

Wearable brain magnetic brain-computer interface system

The invention discloses a wearable brain magnetic brain-computer interface system which can perform high-precision source imaging on brain magnetic signals, improve the spatial resolution of a measurement space decoding mode and realize accurate identification and classification of the brain magnetic signals. The system comprises a wearable brain magnetic acquisition device, a signal analyzing and processing system and a multi-mode stimulation presentation device. The wearable brain magnetic acquisition equipment is used for acquiring brain magnetic signals generated by brain nerve activity of a subject and sending the brain magnetic signals to the signal analysis and processing system; the signal analyzing and processing system is used for preprocessing, imaging and decoding the brain magnetic signals to generate decoding information and sending the decoding information to the multi-mode stimulation presentation equipment; and the multi-modal stimulation presentation equipment is used ...

Supply chain financing information processing method and device

The invention discloses a supply chain financing information processing method and device, and the method comprises the steps: receiving commodity order information issued by a client terminal, initiating a supply contract signing request to a supplier server, and receiving supply contract signing information fed back by the supplier server and product information collected in real time; the supply contract signing information and the product information are synchronized to the Internet of Things; submitting a financing credit line application request to a bank server according to the commodity order information; the bank server collects supply contract signing information and product information from the Internet of Things according to the financing credit line application request, and after verification is passed, a financing credit line verification result fed back by the bank server is received; the verification result is generated after the financing credit line is verified by the ...

Method for training feature extraction model and feature extraction method and device

The embodiment of the invention provides a method for training a feature extraction model and a feature extraction method and device. In the method for training the feature extraction model, a training sample set is utilized to circularly execute the following model training process until a training ending condition is met: image data and text data of each current training sample are respectively provided for a current image feature extraction model and a current text vectorization model; obtaining a corresponding image feature and a corresponding text vector of each current training sample; providing a corresponding text vector of each current training sample and a splicing result of the corresponding image feature and the corresponding text vector to a current multi-modal feature fusion model to obtain a text feature and a multi-modal feature of each current training sample; determining a first loss value and a second loss value according to the multi-modal features and the text features ...