Method and apparatus for processing an image comprising characters

Method and apparatus for processing an image including a character are disclosed. The method may include: searching in a set of characters one or more characters having highest similarities of shape to a character in the set of characters, hereinafter the character being referred to as a first character, the one or more searched characters forming a similar character list of the first character; searching in the set of characters one or more characters having highest similarities of shape to each character in the similar character list of the first character, to form a similar character list of each character in the similar character list of the first character; and selecting in the similar character lists one or more characters having a high mutual similarity between each other, as a character cluster.

Image processing system and method

A method dividing an image into plural superpixels of plural pixels of the image. The method calculates an initial set of weights from a measure of similarity between pairs of pixels, from which a resultant set of weights is calculated for pairs of pixels that are less that a threshold distance apart on the image. The calculation calculates a weight for a pair of pixels as the sum over a set of third pixels of the product of initial weight of the first pixel of the pair of pixel with the third pixel and the weight of the third pixel with the second pixel. Each weight is then subjected to a power coefficient operation. The resultant set of weights and the initial set of weights are then compared to check for convergence. If the weights converge, the converged set of weights is used to divide the image into superpixels.

Image concealing via efficient feature selection

A novel technique for unsupervised feature selection is disclosed. The disclosed methods include automatically selecting a subset of a feature of an image. Additionally, the selection of the subset of features may be incorporated with a congealing algorithm, such as a least-square-based congealing algorithm. By selecting a subset of the feature representation of an image, redundant and/or irrelevant features may be reduced or removed, and the efficiency and accuracy of least-square-based congealing may be improved.

System and method for document analysis, processing and information extraction

The present invention is directed to a method and computer system for representing a dataset comprising N documents by computing a diffusion geometry of the dataset comprising at least a plurality of diffusion coordinates. The present method and system stores a number of diffusion coordinates, wherein the number is linear in proportion to N.

Method and appartaus for data efficient semantic segmentation

A method and system for training a neural network are provided. The method includes receiving an input image, selecting at least one data augmentation method from a pool of data augmentation methods, generating an augmented image by applying the selected at least one data augmentation method to the input image, and generating a mixed image from the input image and the augmented image.

Tda enhanced nearest neighbors

A method comprises receiving a network of a plurality of nodes and a plurality of edges, each of the nodes of the plurality of nodes comprising members representative of at least one subset of initial data points, selecting a subset of the data points based on each node of the plurality of nodes, for each selected data point of the set of selected data points, determining a predetermined number of other data points that are closest in distance to that particular selected data point, grouping the selected data points into a plurality of groups based, at least in part, on the predetermined number of other data points of the set of selected data points that are closest in distance, each group of the plurality of groups including a different subset of data points, and providing a list of selected data points and the plurality of groups.

Hybrid data reduction

An information handling system may include at least one processor and a memory coupled to the at least one processor. The information handling system may be configured to receive data comprising a plurality of data chunks; perform deduplication on the plurality of data chunks to produce a plurality of unique data chunks; determine a compression ratio for respective pairs of the unique data chunks; determine a desired compression order for the plurality of unique data chunks based on the compression ratios; combine the plurality of unique data chunks in the desired compression order; and perform data compression on the combined plurality of unique data chunks.

Multi-domain monitoring services for intelligent infrastructure automation

Data points are collected from multiple infrastructure domains of at least one data center in a computing environment. A graphical representation is generated from the collected data points that represents an infrastructure state of the at least one data center, wherein each vertex in the graphical representation corresponds to one of one or more compute components, one or more storage components, and one or more network components of the data center, and an edge connecting two vertices corresponds to a relationship between the components to which the vertices correspond. The graphical representation representing the infrastructure state of the at least one data center is sent to an infrastructure generation engine for composition with one or more other graphical representations respectively representing the infrastructure state of one or more other data centers in the computing environment.

Systems and methods for identifying electronic content using video graphs

Systems and methods are provided for identifying and recommending electronic content to consumers. In accordance with an implementation, one or more elements of electronic content are identified based on video graph data. In an exemplary method, information associated with a first element of video content is received, and corresponding video graph data is obtained. One or more second elements of video content that are similar to the first element of video content are identified based on the obtained video graph data. A subset the first and second elements of video content is subsequently identified for delivery to the user.

Using machine learning to discern relationships between individuals from digital transactional data

A method including receiving a data structure describing transactions between electronic user accounts associated with users. A relationship graph is constructed from the data in the data structure. The relationship graph has nodes representing entities described in the transactions. The relationship graph has edges representing connections between the nodes. The method also includes clustering groups of nodes within the nodes to form clusters among the nodes. The edges are labeled as relationships types. Labeling is performed by receiving, as input to a machine learning model, a vector having attributes representing the clusters, the nodes, and the edges. Labeling is also performed by outputting, from the machine learning model, probabilities. Each of the probabilities corresponds to a corresponding probability that an edge in the edges represents a relationship type between two nodes in the nodes. Labeling is also performed by labeling, based on the output, the edges as the relationship types.

Video segmentation techniques

A video segmentation system can be utilized to automate segmentation of digital video content. Features corresponding to visual, audio, and/or textual content of the video can be extracted from frames of the video. The extracted features of adjacent frames are compared according to a similarity measure to determine boundaries of a first set of shots or video segments distinguished by abrupt transitions. The first set of shots is analyzed according to certain heuristics to recognize a second set of shots distinguished by gradual transitions. Key frames can be extracted from the first and second set of shots, and the key frames can be used by the video segmentation system to group the first and second set of shots by scene. Additional processing can be performed to associate metadata, such as names of actors or titles of songs, with the detected scenes.

Video segmentation techniques

A video segmentation system can be utilized to automate segmentation of digital video content. Features corresponding to visual, audio, and/or textual content of the video can be extracted from frames of the video. The extracted features of adjacent frames are compared according to a similarity measure to determine boundaries of a first set of shots or video segments distinguished by abrupt transitions. The first set of shots is analyzed according to certain heuristics to recognize a second set of shots distinguished by gradual transitions. Key frames can be extracted from the first and second set of shots, and the key frames can be used by the video segmentation system to group the first and second set of shots by scene. Additional processing can be performed to associate metadata, such as names of actors or titles of songs, with the detected scenes.

Magnetic resonance image analysis method and method for evaluating the risks of radiotherapy

The present disclosure provides a magnetic resonance (MR) image analysis method for a patient who underwent radiotherapy. The method includes the steps: receiving an MR image set of a patient and a dose map of a radiotherapy plan; converting the dose intensity distribution of the dose map into the relative spatial positions in the MR image set; selecting a radiation dose and a radiation exposure region, wherein the radiation exposure region has radiation intensity being equal to or higher than the radiation dose; using the radiation exposure region to determine a region of interest (ROI) in the MR image set; classifying the voxels inside the ROI of the MR image set into different clusters according to the grayscale values of the voxels inside the ROI; and calculating the volume or ratios of the different clusters inside the ROI. The present disclosure also provides a method for evaluating risks of radiotherapy.

Classification of high dimensional data

A method for classification of high dimensional data on graphs based on the Ginzburg-Landau functional. The method applies L 2 gradient flow minimization of the Ginzburg-Landau diffuse interface energy functional to the case of functions defined on graphs. The method performs binary segmentations in a semi-supervised learning (SSL) framework and multiclass tasks are solved by recursively applying a sequence of binary segmentations. Examples illustrate the versatility of the methods on a variety of datasets including congressional voting records, high dimensional test data, and machine learning in image processing.

Assigning privileges in an access control system

An access control system may include a log data parser that receives log data observations in a cloud system and extract user-permission data from the log data observations. The system may also include a clustering unit that uses the user-permission data to generate one or more clusters, each cluster associated with one or more users. Alternatively, and/or additionally, the system may include a feature extractor and a classifier. The feature extractor may extract one or more features from the user-permission data. The classifier may generate predictions of permissions for the one or more users based on the extracted one or more features. The system may also include a policy generator that uses the output of the clustering unit and/or the classifier to generate an access control policy. The policy may be executed in the cloud system to control user's access to one or more services of the system.

Identifying and labeling fraudulent store return activities

A method and system for identifying and labeling fraudulent store return activities includes receiving, by a server, retailer events from an online transaction system of a retailer, the retailer events comprising records of transactions between customers and the retailer, including sale, exchange and return activities across multiple stores. The retailer events are processed to build a network that associates stores, transactions, payment instruments, and customer identification over related activity sequences of transactions. Return fraud labels are generated for the retailer events representing returns based on identified fraud characteristics of the related activity sequences by: representing behavior variables extracted from activity sequences by respective signature vectors; calculating pairwise similarity between the signature vectors; identifying clusters of the signature vectors having common behavior patterns based on the calculated pairwise similarity; and labeling the identified clusters of signature vectors as non-fraudulent behavior or fraudulent behavior.

Hyperspectral naval target detection

The system and method of detecting targets using a camera comprising narrow band filters in a compact pixel cluster arrangement. In some cases, the target is a chemical target. In some cases, the target is a naval target and the processing of the data provides extent, shape, direction and other characteristics that can provide details about the type of naval target even in dark or low light conditions.

Machine learning based application discovery method using networks flow information within a computing environment

A feature selection methodology is disclosed. In a computer-implemented method, components of a computing environment are automatically monitored, and have a feature selection analysis performed thereon. Provided the feature selection analysis determines that features of the components are well defined, a clustering of the features is performed. Provided the feature selection analysis determines that features of the components are well defined, a similarity analysis of the sub-features of the feature is performed. Results of the feature selection methodology are generated.

Data generation device and data generation method

A data generation device according to an embodiment includes one or more processors. The one or more processors classify each of a plurality of users having attributes similar to one another into any one of a plurality of clusters based on user information representing the respective attributes of the users. The one or more processors estimate whether a search result is useful to each of the users based on a similarity between information included in the search result and information representing attributes of the clusters, and output learning data of a ranking model for determining an order of a plurality of the search results, the learning data being data including the search results and estimation information representing whether the search results are useful.

Random walks to detect dissimilar records

One example method includes identifying dissimilar items in a data set. A data set may be walked one or more times and the nodes or vertices of the data set may be scored based on the number of times the nodes are touched during the walks. Scores below a threshold score are determined to be dissimilar nodes in the data set. This allows a diverse set of nodes to be identified. A dissimilar data set may be used to prevent unintentional bias in algorithmic training.

Object recognition using a convolutional neural network trained by principal component analysis and repeated spectral clustering

A method of object recognition trains a convolutional neural network (CNN) with a set of training images, then classifies an image of an object using the trained CNN. A first layer of the CNN is trained by generating a set of first convolutional filters from eigenvectors produced from linear principal component analysis of patches of the training images. The training of each of multiple hidden layers CNN includes generating a set of convolutional filters from a selected subset of eigenvectors produced from linear principal component analysis of patches of an affinity matrix constructed using a set of prior convolutional filters from a prior layer of the CNN, where the affinity matrix represents correlations of feature vectors associated with the prior layer. The last layer of the CNN is trained with a regular classifier by error back-propagation using the training images and labels associated with the training images.

Image processing apparatus and method

There are provided an apparatus and method for generating a heat map that displays crowd density with visual properties such as colors. The image processing apparatus includes at least one processor to implement: an information obtainer configured to obtain an image by photographing a region of interest and location information indicating a location from which the image is photographed; a density map generator configured to generate a crowd density map of the image based on crowd density information estimated from the image; a perspective map generator configured to generate a perspective map that provides distance information indicating a real-life distance between two points of the image based on the location information; and a heat map generator configured to generate a heat map that displays the crowd density information on a map of the region of interest based on the distance information.

Tree delta encoding

Similarity between files nodes of two data tree may be classified based on file names, file paths, and file values. Similarity between directory nodes of two data trees. Responsive to similarity between a file node of a data tree being classified within a no match level: (1) a matching file node of the other data tree may be identified based fingerprints, and (2) a file-node delta may be determined between the file node and the matching file node. A delta between the two data trees may be generated based on the classification of similarity between the file nodes, the classification of similarity between the directory nodes, and the file-node delta.

Optical coin discrimination systems and methods for use with consumer-operated kiosks and the like

Systems and associated methods for optical coin discrimination are disclosed herein. In one embodiment, a method for discriminating coins includes obtaining a digital image of a coin, identifying the outline of the coin using a suitable algorithm, and determining the diameter of the coin using the outline of the coin. The method of this embodiment further includes generating a rectangular image of the coin using, for example, a log-polar transform, generating a series of, for example, Fourier transforms from the rectangular image, and identifying spectral peak locations and intensities in the Fourier transform results. The diameter of the coin, spectral peak location, spectral peak intensity of the coin and/or other aspects of the coin can then be compared to known values for different coins to discriminate the coin.

Image area extraction method and storage program

Provided is an image area extraction method for extracting an image area of an object from a color image of obtained color image data. The image data extraction method includes converting RGB values of each pixel in the color image data into HSV values, performing threshold processing to binarize at least one of the converted S and V values of each pixel so that it will be converted into HS′V′ values, generating composite image data including an X value, a Y value, and a Z value for each pixel, the X value, the Y value, and the Z value being obtained by adding values according to predetermined one-to-one combinations between any one of an R value, a G value, and a B value and any one of an H value, an S′ value, and a V′ value, and extracting the image area using the composite image data.

Habit improving device, method and recording medium

In order to take into account both the user's activity amount and factors other than activity amount to guide the user to have suitable lifestyle habits, this habit improving device is provided with: a first calculation unit which, for every first period, acquires an activity amount of the user totaled for each first period, and which calculates a distribution pattern that indicates the change over time in the activity amount during a second period, which includes a first period; a sorting unit which classifies multiple distribution patterns into one or more groups; a second calculation unit which, on the basis of factors other than activity amount, calculates, and associates with the groups, a target pattern as a target for the user; and a presenting unit which presents advice information to the user on the basis of the current distribution pattern and the target pattern.

Digital picture frames and methods of frame setup

A picture frame and methods of setup, gifting, and/or use. Network connection allows digital frames to be set up remotely by a first user for a second user. The first user can upload photos from electronic devices or from photo collections of community members before the second user receives the frame device. The frame is thus ready for display upon powering on by the second user. An integrated camera is used to automatically determine an identity of a frame viewer and can capture gesture-based feedback. The displayed photos are automatically shown and/or changed according to the detected viewers. The photos can be filtered and cropped at the receiver side. Clustering photos by content is used to improve display and to respond to photo viewer desires.

Automated graph embedding recommendations based on extracted graph features

Embodiments of the invention are directed to a computer-implemented method for matching a graph-under-analysis to a technique for embedding the graph-under-analysis. In a non-limiting example, the computer-implemented method includes receiving, using a processor, graph data representing the graph-under-analysis, wherein the graph-under-analysis represents a network. The graph data is analyzed, using the processor, to extract graph property data representing properties of the graph-under-analysis. Based at least in part on a result of analyzing the graph property data, one or more embedding techniques are selected, wherein at least one of the one or more embedding techniques is configured to transform the graph data to a graph embedding that is used by a task algorithm to perform a task.

Method and system for image segmentation using a directed graph

Method and system is disclosed for image segmentation. The method includes acquiring a digital image, constructing a directed graph from the digital image, calculating a plurality of cost functions, constructing an electrical network based upon the constructed directed graph and the plurality of calculated cost functions, simulating the electrical network using fixed-point linearization, and thresholding the voltages in the simulated electrical network to produce image segmentation. Simulation may be executed in parallel to achieve desirable computational efficiencies.

Co-movement-based automatic categorization system using life-logging data and method thereof

Provided is a co-movement-based automatic categorization method, which includes: receiving image data created by photographing a plurality of subjects; detecting movement data including information on subjects' behaviors from the image data; determining whether or not co-movement between the subjects has been done during each predetermined time interval, and generating co-movement clusters in synchronized time sections, and non-co-movement clusters in desynchronized time sections depending on results of determination; if the duration time of a co-movement cluster is shorter than a standard co-movement duration time, converting the co-movement cluster to a non-co-movement cluster; categorizing the co-movement clusters by grouping a plurality of co-movement clusters existing in the same timeslot into one group; and forming a social network with a co-movement cluster group or a single co-movement cluster, and issuing a social identification to the social network.

Neuropsychological spatiotemporal pattern recognition

Systems and methods for identifying and analyzing neuropsychological flow patterns, include creating a knowledge base of neuropsychological flow patterns. The knowledge base is formed by obtaining signals from multiple research groups for particular behavioral processes, localizing sources of activity participating in the particular behavioral processes, identifying sets of patterns of brain activity for the behavioral processes and neuropsychologically analyzing the localized sources and the identified patterns for each of the research groups. The neuropsychological analysis includes identifying all possible pathways for the identified sets of patterns, ranking the possible pathways based on likelihood for the particular behavioral process and reducing the number of ranked possible pathways based on additional constraints. A system for comparison of obtained signals from an individual to the created knowledge base is provided. These obtained signals are then used to further update the existing knowledge base.

Sample data analysis method based on genomic module network with filtered data

Provided is a method of analyzing sample data based on a genomic module network by means of a computer apparatus. The method includes filtering first gene expression data for a normal or tumor tissue, which is the same tissue as a specific tissue, and second gene expression data for a target tissue to be analyzed, which is the same tissue as the specific tissue, on the basis of a specific module among a plurality of genomic modules; and classifying genes into a plurality of new genomic modules on the basis of an entropy determined using the filtered first gene expression data and determining, for genes belonging to at least one of the plurality of new genomic modules, a first degree of variation of the target tissue relative to the normal or tumor tissue in the at least one genomic module using the filtered first gene expression data and the filtered second gene expression data.

Identifying multiple causal anomalies in power plant systems by modeling local propagations

A system identifies multiple causal anomalies in a power plant having multiple system components. The system includes a processor. The processor constructs an invariant network model having (i) nodes, each representing a respective system component and (ii) invariant links, each representing a stable component interaction. The processor constructs a broken network model having (i) the invariant network model nodes and (ii) broken links, each representing an unstable component interaction. The processor ranks causal anomalies in node clusters in the invariant network model to obtain anomaly score results. The processor generates, using a joint optimization clustering process applied to the models, (i) a model clustering structure and (ii) broken cluster scores. The processor performs weighted fusion ranking on the anomaly score results and broken cluster scores, based on the clustering structure and implicated degrees of severity of any abnormal system components, to identify the multiple causal anomalies in the power plant.

Method and apparatus for recognizing object based on vocabulary tree

A method and apparatus for recognizing an object may obtain, from an input image, feature points and descriptors corresponding to the feature points, determine indices of the feature points based on the descriptors, estimate a density distribution of feature points for each of the indices, and recognize an object included in the input image based on the estimated density distribution.

Counter rare training date for artificial intelligence

A system for enhancing a classifier prediction in respect to underrepresented classes may be provided. A classifier system trained with training data to build a model is used for classifying unknown input data, and an evaluator engine adapted for a determination of an underrepresented class. Additionally, the system comprises an extractor engine adapted for an extraction of relating data from an additional source, and a similarity engine adapted for a selection of data sets out of the relating data wherein the similarity engine is also adapted for comparing features of the relating data and a representative data set for the underrepresented class. Finally, the system comprises a recursion unit adapted for triggering the evaluator engine, the extractor engine and the similarity engine treating selected data set as input data until the evaluator engine classifies the selected data set with a confidence level which is above a confidence threshold level.

Digital picture frame photograph clustering

A method for automated routing of pictures taken on mobile electronic devices to a digital picture frame including a camera integrated with the frame, and a network connection module allowing the frame for direct contact and upload of photos from electronic devices or from photo collections of community members. The integrated camera is used to automatically determine an identity of a frame viewer and can capture gesture-based feedback. The displayed photos are automatically shown and/or changed according to the detected viewers. The photos can be filtered and cropped at the receiver side. Clustering photos by content is used to improve display and to respond to photo viewer desires.

Image photographing apparatus and control method thereof

영상 촬영 장치 및 이의 제어 방법이 제공된다. 본 영상 촬영 장치의 제어 방법은 사용자의 촬영 명령이 입력되면, 기설정된 시간 내에 연속 촬영을 수행하여 초점 거리가 동일한 복수의 영상을 획득하고, 복수의 영상을 이용하여 모션 벡터를 산출하고, 모션 벡터 및 복수의 영상 중 제1 영상의 색 정보를 바탕으로 제1 영상의 전경 및 후경을 분리하며, 분리된 전경 및 후경을 바탕으로 아웃 포커싱을 수행한다. An image photographing apparatus and a method for controlling the same are provided. In this method of controlling an image photographing apparatus, when a user's photographing command is input, continuous photographing is performed within a preset time to acquire a plurality of images having the same focal length, a motion vector is calculated using the plurality of images, and a motion vector and separating a foreground and a rear view of the first image based on color information of the first image among the plurality of images, and performing out-focusing based on the separated foreground and rear view.

Multiple-wavelength images analysis electro optical system for detection of accident ship and submerged person and analysis method thereof

A multi-wavelength image analysis electro-optical system for detecting accident ships and drowned people according to the present invention is a multi-wavelength image analysis electro-optical system for detecting accident ships and drowned people, the system comprising an input unit (10) comprising a super-low illumination camera (3), a short-wavelength infrared image sensor (5), a medium-wavelength infrared image sensor (7), and a long-wavelength infrared image sensor (9); a signal processing unit (20) which receives and processes data of the input unit (10); a display unit (30) which receives and displays data of the signal processing unit (30); a storage unit (40) which stores data of the signal processing unit (20) and the display unit (30); and a control unit (90) which has a camera control board (50) and a driving control board (60) and controls the input unit (10), the signal processing unit (20), the display unit (30), and the storage unit (40).

Lane line processing method and device

本发明实施例提出一种车道线处理方法和装置。该方法包括：对第一图像进行二值化处理，得到二值图像，所述第一图像中包括车道线点和非车道线点；对所述二值图像进行连通域分析，得到所述二值图像中的一个或多个连通域，每个连通域中包括相互邻接的多个车道线点；利用各所述连通域确定各车道线对应的分组中包括的车道线点；利用各所述分组中包括的车道线点得到各所述分组对应的车道线的表示信息。本发明实施例，利用连通域分析方法对二值图像上的车道线点进行分组，得到的分组准确，不受图像质量的影响，具有较高的鲁棒性。

Image topological coding for visual search

A method and an apparatus for processing an image generate a first vector of a first number dimension for the image from a first number of points of the image based on topological information of the first number of points, and the first vector for the image is invariant to rotation and scaling in creating the image. The first number of points may be locations of a set of rotation and scaling invariant feature points for the image, and the generated first vector may be a graph spectrum of a pair-wise distance matrix generated from the first number of points of the image.

Picture identification method and device, electronic equipment and medium

本发明实施例公开了一种图片识别方法，在通过光学字符识别模型对采集的图像进行文字识别过程中，获取识别出的当前字符串及其哈希值，并将当前字符串及其哈希值存储到第一预设树形结构及第二预设树形结构中中，预测当前字符串在下一时刻的新概率值，获取扩充字符串集，根据第二预设树形结构中存储的哈希值，获取概率值最高的N个字符串并保留，并将N个字符串作为当前字符串，重复上述步骤直至完成所有采集的图像的识别，获取概率值最高的一个字符串作为最终识别结果。本发明实施例提供的图片识别方法、装置、电子设备及介质，能够有效降低针对图像进行文字识别过程存储中间字符串所需的计算量，能够有效提高存储中间字符串的效率。

Optical coin discrimination systems and methods for use with consumer-operated kiosks and the like

Systems and associated methods for optical coin discrimination are disclosed herein. In one embodiment, a method for discriminating coins includes obtaining a digital image of a coin, identifying the outline of the coin using a suitable algorithm, and determining the diameter of the coin using the outline of the coin. The method of this embodiment further includes generating a rectangular image of the coin using, for example, a log-polar transform, generating a series of, for example, Fourier transforms from the rectangular image, and identifying spectral peak locations and intensities in the Fourier transform results. The diameter of the coin, spectral peak location, spectral peak intensity of the coin and/or other aspects of the coin can then be compared to known values for different coins to discriminate the coin.

Image processing apparatus, control method therefor, and program

Methods and systems for effective automatic recognition of symbols using forest solutions

FIELD: optics. SUBSTANCE: invention relates to optical character recognition. Proposed system includes instructions in machine code, when executed by a processor, control system of optical character recognition to process scanned image containing text of document by performing identification of character images containing text in scanned image of document. Identification is performed for each document page and symbol for each image on page. Method includes identification of a set of suitable reference data structures for image of symbol using forest solutions. Method uses a suitable standard data structure to determine appropriate set of grapheme and uses identified set of suitable grapheme to select character code that corresponds to image of characters. Method includes preparing processed document containing codes symbols, which correspond to images of characters from a scanned document image, and processed document in one or more memory devices and memory. EFFECT: optimising optical character recognition owing to use of forest solutions. 20 cl, 66 dwg РОССИЙСКАЯ ФЕДЕРАЦИЯ (19) RU (11) (13) 2 582 064 C1 (51) МПК G06K 9/68 (2006.01) G06F 17/21 (2006.01) G06K 9/62 (2006.01) G06K 9/80 (2006.01) ФЕДЕРАЛЬНАЯ СЛУЖБА ПО ИНТЕЛЛЕКТУАЛЬНОЙ СОБСТВЕННОСТИ (12) ОПИСАНИЕ (21)(22) Заявка: ИЗОБРЕТЕНИЯ К ПАТЕНТУ 2014150945/08, 16.12.2014 (24) Дата начала отсчета срока действия патента: 16.12.2014 (72) Автор(ы): Чулинин Юрий Георгиевич (RU), Сенкевич Олег Евгеньевич (RU) (45) Опубликовано: 20.04.2016 Бюл. № 11 2 5 8 2 0 6 4 R U (54) СПОСОБЫ И СИСТЕМЫ ЭФФЕКТИВНОГО АВТОМАТИЧЕСКОГО РАСПОЗНАВАНИЯ СИМВОЛОВ С ИСПОЛЬЗОВАНИЕМ ЛЕСА РЕШЕНИЙ (57) Реферат: Изобретение относится к оптическому идентификацию набора подходящих структур распознаванию символов. Техническим данных эталона для изображения символа с результатом является оптимизация оптического использованием леса решений. Используют распознавания символов за счет использования подходящие структуры данных эталона для леса решений. ...

High-spectral data dimension reduction method based on rarefaction representation and empty spectrum Laplce's figure

本发明公开了一种用于大规模高光谱数据的降维方法，主要用于解决传统流行学习信息单一和该类方法难以处理较大规模数据的问题。其实现步骤为：1.从大规模的高光谱数据中选择一定量的数据作为训练样本；2.对训练样本进行空谱拉普拉斯图的构造；3.对拉普拉斯矩阵进行特征分解得到训练样本的低维表示；4.利用训练样本及其低维表示构造高维字典和低维字典；5.计算剩余高光谱数据在高维字典上的稀疏表示系数；6.将该稀疏表示系数与低维字典进行相乘，得到剩余数据的低维表示；7.整合训练样本与剩余数据的低维表示得到完整的降维数据。本发明提升了流行降维的效果，可用于处理大规模的高光谱数据。

Adaptive non-intrusion type load recognition methods based on random forest

基于随机森林的自适应非侵入式负荷识别方法，包括建立用电负荷特征数据库；从每个切换事件中提取所需的负荷特征；将所得负荷特征进行归一化处理后，获得所需样本点；样本点由未知模式识别模块处理，为样本点分配“已知”标签、或“未知”标签；所有标签为“已知”样本点，利用随机森林算法得出识别结果；所有标签为“未知”样本点，由在线聚类模块处理，若产生新聚类，用户可以选择是否将类标签分配给聚类；带标签的新聚类，通过在线更新模块更新随机森林和未知模式识别模块对已有知识进行更新;“未知点”通过新随机森林得出识别结果。本发明可以将容易错误分类的负荷识别为“未知”，并在获取新知识后完成正确的识别，有利于促进未知负荷模式的有效识别。

METHOD FOR GROUPING IMAGES OF A VIDEO SEQUENCE

Automatic screening method of group sharing photos

提出了一种群共享照片的自动筛选方法，属于图像处理技术领域。包括步骤S1，创建摄影者因子表、比例因子表和位置因子表，所述摄影者因子表包括摄影者标识、摄影者相机指纹以及摄影因子三个表项，所述比例因子表包括摄像主体所占比例以及比例因子两个表项，所述位置因子表包括摄像主体所在位置以及位置因子两个表项；步骤S2，从处于群中各个用户接收共享的多幅图像，形成初始图像集合；步骤S3，对初始图像集合进行筛选；步骤S4，输出筛选出的图像，所述筛选出的图像为质量值高的图像。本方案能够免除用户过多繁复的操作，节省用户的时间和精力，自动从群共享的照片中挑选出优质照片并显示给群用户，节省了用户时间、网络资源以及存储资源。

Semi-supervised multi-spectral remote sensing image segmentation method based on spectral clustering

本发明公开了一种基于谱聚类的半监督多光谱遥感图像分割方法，其分割过程是：(1)提取输入多光谱遥感图像的特征；(2)对具有S个像素点的多光谱遥感图像，随机均匀采样N个无标签的点和M个有标签的点的集合 ，n＝N+M，其中M个有标签的点用来构造成对限制信息Must-link和Cannot-link集合；(3)对采样的点集合 ，采用半监督的谱聚类进行分类，得到这n＝N+M个点的类别标签；(4)将采样的n＝N+M个点作为训练样本，对剩余的(S-N-M)个点用最近邻的准则进行分类，为每个像素点按其所在的类别赋一个类标，作为输入图像的分割结果。本发明与现有技术相比分割效果良好，可操作性强，分类精度有所提高，避免反复试验寻找最优的参数，对图像大小限制小，可更好的应用于多类别的多光谱遥感图像的分割。

USE OF SILHOUETTES FOR RAPID RECOGNITION OF OBJECTS

Un procédé de reconnaissance d'objets comprenant les étapes d'obtention d'une image à partir d'un capteur d'image et d'un nuage de points 3D à partir d'un capteur de profondeur ; de synchronisation de l'image et du nuage de points 3D ; de groupement des points 3D pour séparer les objets du nuage de points 3D ; d'extraction de silhouettes par la segmentation de l'image en utilisant le groupement de points 3D ; et de détection des contours des objets séparés dans l'image segmentée ; de reconnaissance de silhouettes par la transformation de chaque contour détecté en un descripteur de silhouettes, et de classification de ces descripteurs de silhouettes en des objets reconnus en utilisant un réseau neuronal formé pour la reconnaissance d'objets. An object recognition method comprising the steps of obtaining an image from an image sensor and a 3D point cloud from a depth sensor; synchronization of the image and the 3D point cloud; grouping 3D points to separate objects from the 3D point cloud; extraction of silhouettes by segmentation of the image using the grouping of 3D points; and detecting the outlines of the separated objects in the segmented image; of recognizing silhouettes by transforming each detected contour into a silhouettes descriptor, and classifying these silhouetted descriptors into recognized objects using a neural network formed for the recognition of objects.

A kind of improved spectral clustering and parallel method

本发明公开了一种基于群智能算法的改进谱聚类方法，通过选取拉普拉斯矩阵的前2k个最大特征值对应的特征向量作为聚类的源数据，然后通过群智能算法选取优质的初始化中心点进行聚类操作，提高了聚类的最高准确率和多次聚类结果的稳定性。本发明引入了布谷鸟搜索算法寻找初始化中心点，将布谷鸟搜索算法过程中的适应度函数采用误差平方和函数，应用于谱聚类中，将搜索得到的最小的误差平方和的数据点作为初始化中心点。本发明将布谷鸟搜索算法中的莱维飞行策略引入粒子群算法，在粒子群算法收敛速度变缓情况下，使用莱维飞行策略产生频繁较小步长和和偶尔较大步长，在不同步长下引入不同侧重的速度更新公式。

Method for reconstructing a 3D scene as a 3D model using images acquired by 3D sensors and omnidirectional cameras

A method reconstructs a three-dimensional (3D) scene as a three-dimensional (3D) model by first acquiring red, green, blue, and depth (RGBD) images with a sensor, and wide field-of-view (FOV) red, green, and blue (RGB) images with a camera from different viewpoints. Pair-wise point matches among the RGBD images and the RGB images are identified, and a graph of nodes and edges is constructed using the pair-wise point matches. Initial poses of the RGBD images and the RGB images are estimated using pose constraints associated with the edges, and an initial 3D model of the scene is estimated using the estimated poses. Then, erroneous edges are removed in the graph, and bundle adjustment is applied to the initial poses and the initial 3D model to obtain refined poses and the 3D model of the scene.

Method and apparatus for processing an image comprising characters

Method and apparatus for processing an image including a character are disclosed. The method may include: searching in a set of characters one or more characters having highest similarities of shape to a character in the set of characters, hereinafter the character being referred to as a first character, the one or more searched characters forming a similar character list of the first character; searching in the set of characters one or more characters having highest similarities of shape to each character in the similar character list of the first character, to form a similar character list of each character in the similar character list of the first character; and selecting in the similar character lists one or more characters having a high mutual similarity between each other, as a character cluster.

Ranking of images and image labels

The subject matter of this specification can be embodied in, among other things, a method that includes determining a score for an image of a plurality of images with respect to each of one or more terms, identifying one or more of the terms for each of which the score for the image with respect to the respective identified term satisfies a criterion, and associating the identified terms with the image. Determining the score for the image with respect to a respective term includes determining probabilities of navigating between images in the plurality of images and determining the score for the image with respect to the respective term based on the probabilities.

Automated identification and use of building floor plan information

Techniques are described for using computing devices to perform automated operations for identifying building floor plans that have attributes satisfying target criteria and for subsequently using the identified floor plans in further automated manners. In at least some situations, the identification of such building floor plans is based on generating and using adjacency graphs generated for the floor plans that represent inter-connections between rooms and other attributes of the buildings, and in some cases is further based on generating and using embedding vectors that concisely represent the information of the adjacency graphs. Information about such identified building floor plans may be used in various automated manners, including for controlling navigation of devices (e.g., autonomous vehicles), for display on client devices in corresponding graphical user interfaces, for further analysis to identify shared and/or aggregate characteristics, etc.

Method and apparatus for processing an image comprising characters

Method and apparatus for processing an image including a character are disclosed. The method may include: searching in a set of characters one or more characters having highest similarities of shape to a character in the set of characters, hereinafter the character being referred to as a first character, the one or more searched characters forming a similar character list of the first character; searching in the set of characters one or more characters having highest similarities of shape to each character in the similar character list of the first character, to form a similar character list of each character in the similar character list of the first character; and selecting in the similar character lists one or more characters having a high mutual similarity between each other, as a character cluster.

Multiple-instance-learning-based video classification

A classifier training system trains unified classifiers for categorizing videos representing different categories of a category graph. The unified classifiers unify the outputs of a number of separate initial classifiers trained from disparate subsets of a training set of media items. The training process divides the training set into a number of bags, and applies a boosting algorithm to the bags, thus enhancing the accuracy of the unified classifiers.

Network killing chain detection method, prediction method and system

本发明公开了一种网络杀伤链检测方法、预测方法及系统，具体包括：(1)构建d维特征向量；(2)无监督特征选择算法将d维特征向量筛减为k维；(3)通过k维特征向量获取网络杀伤链攻击事件序列集合。在IDS告警日志数据进行杀伤链挖掘的真实场景中，针对无法提前知晓数据中所包含杀伤链数目的问题，本发明改进的谱聚类算法相比于其他的有监督学习方法不仅能够实现无监督学习，还能够自动识别聚类数目；(4)基于已经获得的网络杀伤链序列，采用马尔科夫理论与三种网络杀伤链变种模型进行预测分析；(5)基于理论分析，实现了杀伤链检测与预测系统。

Analysis method, analysis program and analysis system using graph theory

Key frame is identified using the openness analysis of group

本发明涉及一种用于从视频序列识别关键视频帧集合的方法，其包括提取每一视频帧的特征向量及应用群组稀疏性算法以将特定视频帧的所述特征向量表示为其它视频帧的所述特征向量的群组稀疏组合。分析与所述群组稀疏组合相关联的加权系数以确定在时间上邻接的类似视频帧的视频帧群集。基于所述所确定视频帧群集选择关键视频帧集合。

Rationalizing network predictions using similarity to known connections

Rationalization of network predictions using similarity to known connections is provided. In various embodiments, a graph is read. The graph comprises a plurality of nodes. Each of the plurality of nodes corresponds to an entity or property. The plurality of nodes is interconnected by a plurality of edges. Each edge corresponds to a relationship between connected nodes. A new edge in the graph is predicted. The new edge corresponds to a relationship between a first node and a second node. The first node corresponds to an entity and the second node corresponds to an entity or property. One or more additional nodes connected to the second node is located. The one or more additional nodes is scored according to its connections in common with the first node. One or more sources is provided to a user describing the connection between the one or more additional node and the second node.

Method and apparatus for constructing decision model, computer device and storage device

A method of constructing a decision model includes: obtaining a rule template data and extracting each variable object and each template sample from the rule template data; clustering and analyzing the variable objects to obtain a clustering result; matching the clustering result with each template sample according to the rule template data, and serving the matched clustering result as a first feature; calculating a black sample probability for each variable object and serving the black sample probability of each variable object as a second feature; and constructing the decision model according to the first feature and the second feature.

A kind of multimode dispatching method, apparatus and system

本发明公开一种多模块调度方法、装置和系统，以解决现有多模块调度技术中存在的计算模块的输入数据存在不一致的问题。该方法包括：主进程读取预存的配置文件，配置文件中保存有一个有向计算图；主进程在当前计算周期中，在计算开始时初始化有向计算图中全部节点和全部连接边的状态；根据有向计算图中的计算方向以及节点的状态确定待调用节点，待调用节点包括节点的多个输入边均为完成状态的节点；向待调用节点对应的从进程中的计算模块发送RPC调用请求以执行该计算模块；在收到应答响应后，更新该节点的状态和该节点的输出边的状态；在确定全部节点的状态更新结束后，进行下一个计算周期的处理。

Dynamic grid generation method, generation device, polymerization method and polymerization device

本发明公开了一种动态网格的生成方法，包括：将当前显示区域所在的最大、最小经纬度线构成的封闭区域作为所述当前显示区域的参考区域；找出与所述参考区域上任意一点最近的基准点作为最近基准点，所述最近基准点从预存的基准点数据库找出或从已知的基准点即时生成；以所述最近基准点为起始点向经纬度方向延伸，生成由多个所述单元网格构成且完全覆盖所述当前显示区域的最小动态网格。本发明还公开了一种动态网格的生成装置、基于动态网格的聚合方法和聚合装置。采用本发明实施例，可以有效解决网格聚类算法中需存储大量数据、数据处理慢、聚合误差大的问题。

METHOD FOR GROUPING IMAGES OF A VIDEO SEQUENCE

Le procédé exploitant une structure en graphe est caractérisé en ce qu'il comporte l'itération suivante :- calcul du potentiel du noeud nm (8), fusion de deux noeuds ni et nJ, en fonction des distances entre les attributs des images clés et en fonction de la distance temporelle de ces images clés,- calcul du potentiel de chaque arête (8) reliant le noeud fusionné à un autre noeud du graphe précédemment relié au noeud ni ou nJ,- fusion des deux noeuds et validation du nouveau graphe (10) si l'énergie de ce graphe est inférieure à l'énergie du graphe avant la fusion (9).Une application concerne la construction automatique de résumés vidéo. The method exploiting a graph structure is characterized in that it comprises the following iteration: - calculation of the potential of the node nm (8), fusion of two nodes ni and nJ, according to the distances between the attributes of the key images and as a function of the temporal distance of these key images, - calculation of the potential of each edge (8) connecting the merged node to another node of the graph previously connected to the node ni or nJ, - fusion of the two nodes and validation of the new graph ( 10) if the energy of this graph is lower than the energy of the graph before the fusion (9). An application concerns the automatic construction of video summaries.

Determining the position of a mobile device in a geographical area

A method of determining the position of a mobile device using a model. The model comprises a graph comprising a set of vertices and a set of edges between the vertices, and one or more feature functions that take as input one or more vertices of the graph and a set of observations for the device, and return a feature value. The model is constructed by obtaining a map of the geographical area, constructing the graph using the map, wherein the vertices of the graph correspond to areas of the map, and two vertices are joined by an edge when the areas corresponding to the vertices are adjacent and can be travelled between by a device, and building the one or more feature functions using the graph, wherein the feature value of the one or more feature functions indicates the extent to which the observations support the device being positioned in the areas corresponding to the one or more vertices of the graph. The method then comprises the steps of: obtaining a set of observations from the sensors of the device; determining the trajectory for the device that optimises the values given by the feature functions given for the set of observations; and determining the position of the device to be the end point of the trajectory.

Image segmentation method based on characteristic importance sorting spectral clustering

本发明公开了一种基于特征重要度排序谱聚类的图像分割方法，主要解决现有谱聚类方法复杂度高和稳定性差的问题。其实现过程是：(1)对待分割图像提取灰度值特征，灰度共生特征或小波特征；(2)对特征数据进行归一化处理；(3)根据归一化后的特征数据，计算所有样本的重要度；(4)对所有样本的重要度进行排序，并选择重要度高的100个样本作为采样的样本子集；(5)根据 方法，用选出的样本子集求解所有样本谱映射后的特征向量空间，并取前给定的类别数k个特征值对应的特征向量进行降维；(6)对降维后的数据进行k-means聚类，输出最终的图象分割结果。本发明与现有的谱聚类方法相比，结果稳定，复杂度低，图像分割结果有明显提高，可用于目标检测和目标识别。

Image segmentation

Method of grouping images from a video sequence

The method utilizing a graph-like structure is comprises the following iteration: calculation of the potential of node n m , merging of two nodes n i and n j , as a function of the distances between the attributes of the key images and as a function of the temporal distance of these key images, calculation of the potential of each edge connecting the merged node to another node of the graph previously connected to nodes n i or n j , merging of the two nodes and validation of the new graph if the energy of this graph is less than the energy of the graph before merging.

Data-driven text incremental clustering method

本发明公开了一种数据驱动的文本增量聚类方法，属于自然语言处理领域，包括以下步骤：根据已有聚类类别判断是否需要增量聚类，并根据判断结果进行初始化聚类生成聚类类别，或对已有聚类类别进行簇心调整，更新已有聚类类别，实现文本增量聚类。本发明实现对对增量文本的自动聚类，保持已有聚类结果的延续性，提高了文本聚类的准确性。

Content-based and time-evolving social network analysis

System and method for modeling a content-based network. The method includes finding single mode clusters from among network (sender and recipient) and content dimensions represented as a tensor data structure. The method allows for derivation of useful cross-mode clusters (interpretable patterns) that reveal key relationships among user communities and keyword concepts for presentation to users in a meaningful and intuitive way. Additionally, the derivation of useful cross-mode clusters is facilitated by constructing a reduced low-dimensional representation of the content-based network. Moreover, the invention may be enhanced for modeling and analyzing the time evolution of social communication networks and the content related to such networks. To this end, a set of non-overlapping or possibly overlapping time-based windows is constructed and the analysis performed at each successive time interval.

Image clustering using distance, local linear structure and affine symmetry

An adaptive clusterer

This invention relates to a technique and system (34) for clustering a set of data points (18) in a data field. Clustering is performed by a set of rules which make no assumptions about the data fields. That is, the number of clusters, the distance between clusters, etc., need not be specified in advance. The invention calculates a distance R that is a function of the distances between each pair of points (18) in a data field. Operating on a chosen point, a parameter is calculated for each neighboring point within distance R from the chosen point. This parameter is proportional to the distance of the neighboring point to the chosen point. The chosen point P is then connected to a neighboring point on the basis of this parameter. All points, in any order are operating on as the new chosen point, so that the additional points may be linked to new chosen points in a similar manner. In this way, a tree-structure (22, 24) of connected points will form the desired clustering. Points (18) not having other points within distance R will not be clustered or linked.

Lane line treating method and apparatus

本发明实施例提出一种车道线处理方法和装置。该方法包括：对第一图像进行二值化处理，得到二值图像，所述第一图像中包括车道线点和非车道线点；对所述二值图像进行连通域分析，得到所述二值图像中的一个或多个连通域，每个连通域中包括相互邻接的多个车道线点；利用各所述连通域确定各车道线对应的分组中包括的车道线点；利用各所述分组中包括的车道线点得到各所述分组对应的车道线的表示信息。本发明实施例，利用连通域分析方法对二值图像上的车道线点进行分组，得到的分组准确，不受图像质量的影响，具有较高的鲁棒性。

Banded graph cut segmentation algorithms with laplacian pyramids

A process for segmenting an object of interest from background, comprising: obtaining a master, high-resolution image of an object disposed within a background; applying a first band graph cut process to the master image generating a second image with the object being segmented from the background to a first approximation; and, comparing the second image with the master image to produce a comparison image with pixels identified by the comparison to be background images being removed from the comparison image.

Apparatus, system and method for extracting a region of interest from an image using graph cuts

An image processing apparatus comprises generating means (106) for generating a second image by changing a size of a first image such that a region of interest in the second image satisfies a predetermined criterion about a dimension; and extracting means (107) for extracting the region of interest from the second image generated by the generating means, by applying a Graph-Cut method using a Graph-Cut coefficient corresponding to the criterion to the second image.

Method and system for clustering data

A method of classifying a plurality of elements, such as genes, an associated system and an associated computer-readable storage medium. Similarity values for pairs of elements are measured. For example, in the case of genes, gene expression fingerprints are measured, and the similarity values are computed from the fingerprints. A graph is constructed such that each vertex of the graph corresponds to a respective element. Each edge of the graph is assigned a superbinary weight that is based on the corresponding similarity value. The graph is partitioned into kernels, and the kernels are merged into clusters. Preferably, the superbinary weights are based on the similarity values according to a probabilistic model. The system of the present invention includes an apparatus for measuring the similarity values, a memory for storing the similarity values, and a proccesor for implementing the method of the present invention. The storage medium of the present invention includes computer readable code in which the method of the present invention is encoded.

Method and System for Recognizing Target Parking Location

본 발명은 차량의 목표주차위치 인식 방법 및 그 시스템에 관한 것이다. The present invention relates to a method and a system for recognizing a target parking position of a vehicle. 본 발명은, 차량의 목표주차위치 인식 시스템에 있어서, 상기 차량의 후면으로부터 반사되어 입수되는 레인지 데이터를 검출하는 스캐닝 레이저 레이더, 상기 레인지 데이터를 전처리하여 유효 클러스터를 추출하는 전처리부, 상기 유효 클러스터로부터 장애물 코너를 추출하는 코너인식부 및 상기 장애물 코너 중에서 인접공간이 빈주차공간 조건을 만족하되 자차와 가장 인접한 상기 장애물 코너인 메인 레퍼런스 코너를 인식하고 상기 인접공간쪽으로 상기 메인 레퍼런스 코너로부터 차폭 방향의 역방향에 존재하되 차폭거리과 차 길이 사이의 범위에 있는 가장 가까운 상기 유효 클러스터 상의 점인 서브 레퍼런스 코너를 인식하여 목표주차위치를 설정하는 주차위치 설정부를 포함하는 것을 특징으로 하는 목표주차위치 인식 시스템을 제공한다. The present invention relates to a target parking position recognition system for a vehicle, comprising: a scanning laser radar for detecting range data reflected from a rear surface of the vehicle, a preprocessor for preprocessing the range data, and extracting an effective cluster from the effective cluster; A corner recognition unit for extracting an obstacle corner and an adjacent space satisfies the empty parking space condition among the obstacle corners, but recognizes a main reference corner which is the obstacle corner closest to the host vehicle, and reverses the vehicle width direction from the main reference corner toward the adjacent space. And a parking position setting unit configured to set a target parking position by recognizing a sub-reference corner that is present on the nearest valid cluster in the range between the vehicle width distance and the vehicle length. 본 발명에 의하면, 차량 후방에 스캐닝 레이저 레이더를 장착하여 감지한 레인지 데이터를 이용하여 직각 주차시 목표주차위치를 설정함으로써 주간, 야간, 야외, 지하, 역광 등 기존의 다른 방법들에선 해결하기 힘들었던 경우에도 목표주차위치 설정이 가능해지는 효과가 있다. According to the present invention, when it is difficult to solve in the existing methods such as daytime, nighttime, outdoor, underground, backlighting by setting a target parking position at right angle parking using the range data detected by mounting a scanning laser radar at the rear of the vehicle. Also, the target parking position can be set. 주차위치설정, 데이터 클러스터, 빈주차공간  Parking position setting, ...

Forming connections between image collections

A method of establishing a connection between two image collections, the method includes providing a first image collection from a first user; providing a second image collection from a different second user; analyzing the first and second image collections to produce a similarity score between the first and second image collections; and establishing a link for sharing images or collection data between the first and second users based on the similarity score.

Method and apparatus for classifying known specimens and media using spectral properties and identifying unknown specimens and media

Method and apparatus for determining a metric for use in predicting properties of an unknown specimen belonging to a group of reference specimen electrical devices comprises application of a network analyzer for collecting impedance spectra for the reference specimens and determining centroids and thresholds for the group of reference specimens so that an unknown specimen may be confidently classified as a member of the reference group using the metric. If a trait is stored with the reference group of electrical device specimens, then, the trait may be predictably associated with the unknown specimen along with any traits identified with the unknown specimen associated with the reference group.

News group clustering based on cross-post graph

A system and/or method that facilitates analyzing newsgroup clusters. A data reception component receives data relating to a plurality of newsgroups and relays the data to an engine that constructs a weighted graph. The weighted graph represents a subset of the newsgroups as vertices of the graph. The vertices are connected by edges, which represent cross-postings relating to the subset of newsgroups.

An adaptive clusterer

Subspace clustering method and device for potential low-rank representation

本发明公开了一种潜在低秩表示的子空间聚类方法及装置，通过获取数据并对其进行预处理，得到特征矩阵；利用考虑了未观测到的数据样本潜在低秩表示子空间聚类，并利用Schatten‑p范数作为正则项来代替秩函数，将NP难求解的问题转化为一个可求解的问题；引入l p 范数约束误差项以构建潜在低秩表示子空间聚类的优化目标函数；然后求解优化目标函数得到低秩表示矩阵；基于低秩表示矩阵计算亲和矩阵；利用谱聚类算法对亲和矩阵进行计算分割，实现数据的潜在低秩表示子空间聚类。本发明解决了低秩表示样本不足和秩函数难求解的问题，增强了潜在低秩表示子空间聚类的鲁棒性，提升了潜在低秩表示子空间聚类的性能。

Method and device of constructing decision model, computer device and storage apparatus

A method of constructing a decision model includes: obtaining a rule template data and extracting each variable object and each template sample from the rule template data; clustering and analyzing the variable objects to obtain a clustering result; matching the clustering result with each template sample according to the rule template data, and serving the matched clustering result as a first feature; calculating a black sample probability for each variable object and serving the black sample probability of each variable object as a second feature; and constructing the decision model according to the first feature and the second feature.

An image processing system and method

A method dividing an image into plural superpixels of plural pixels of the image. The method calculates an initial set of weights from a measure of similarity between pairs of pixels, from which a resultant set of weights is calculated for pairs of pixels that are less that a threshold distance apart on the image. The calculation calculates a weight for a pair of pixels as the sum over a set of third pixels of the product of initial weight of the first pixel of the pair of pixel with the third pixel and the weight of the third pixel with the second pixel. Each weight is then subjected to a power coefficient operation. The resultant set of weights and the initial set of weights are then compared to check for convergence. If the weights converge, the converged set of weights is used to divide the image into superpixels.

Method for computation relating to clumps of virtual fibers

A computer-implemented method for processing a set of virtual fibers into a set of clusters of virtual fibers, usable for manipulation on a cluster basis in a computer graphics generation system, may include determining aspects for virtual fibers in the set of virtual fibers, determining similarity scores between the virtual fibers based on their aspects, and determining an initial cluster comprising the virtual fibers of the set of virtual fibers. The method may further include instantiating a cluster list in at least one memory, adding the initial cluster to the cluster list, partitioning the initial cluster into a first subsequent cluster and a second subsequent cluster based on similarity scores among fibers in the initial cluster, adding the first subsequent cluster and the second subsequent cluster to the cluster list, and testing whether a number of clusters in the cluster list is below a predetermined threshold.

Directed acyclic graph generation method, apparatus, device, and medium

本申请公开了一种有向无环图的生成方法、装置、设备及介质，属于有向无环图领域。所述方法包括：获取n个业务数据；在目标属性维度上基于属性取值，对n个业务数据进行第1次聚类，得到m 1 个聚类结果；将m 1 个聚类结果设置为有向无环图的m 1 个一级节点，设置初始节点是m 1 个一级节点的父节点；在目标属性维度上基于属性取值，针对m i‑1 个聚类结果中的第j个聚类结果进行第i次聚类，得到与第j个聚类结果对应的m i 个聚类结果；将m i 个聚类结果设置为有向无环图的m i 个i级节点；设置与第j个聚类结果对应的第j个i‑1级节点是m i 个i级节点的父节点；显示有向无环图。上述方法提高了有向无环图的生成效率。

Method for improving classification results of a classifier

A method for improving classification results of a classifier including receiving classification results for a plurality of elements that have been classified by a classifier as one of a plurality of classes, constructing a graph having a plurality of nodes, each node corresponding to one of the elements, and a plurality of labels, each label corresponding to one of the classes, adding edges between nodes corresponding to related elements, adding edges between each node and each label, and using a graph cut algorithm to cut edges to a node and partition the graph into classes, the graph cut algorithm using as input the classification results for the element corresponding to that node and related elements.

Landmark feature selection

An example method comprises receiving a multidimensional data set, receiving a predetermined number of features for a set of landmark features, when a current number of features of the set is less than the predetermined number: for each landmark feature of the set of landmark features, calculate a distance between that particular landmark feature and each non-selected feature that is not within the set, identify a closest non-selected feature to that particular landmark feature, identify a particular closest non-selected feature related to a largest distance among the distances, and adding the particular non-selected feature to the set of landmark features, and if the current number of features of the set of landmark features is equal to or greater than the predetermined number of features for the set of landmark features, then providing identification of at least a subset of features of the set of landmark features.

Method for evaluating link quality by adopting multi-granularity cascade forest

本发明公开了一种采用多粒度级联森林评估链路质量的方法，该方法通过利用无线传感网络的链路特性与多粒度级联森林(gcForest，multi‑Grained Cascade forest)模型相结合，提出了一种无线传感器网络链路质量评估的方法。该方法首先获取实验场景下的样本数据，运用步长和中心点优化的聚类算法对样本数据划分链路质量等级；利用分层抽样获取链路中三个区域的样本进行训练，结合级联森林堆叠结构建立多粒度级联森林链路质量评估模型；最后通过准确性、稳定性和敏捷性评价整个链路质量评估模型的性能。本发明优点在于运用链路区域特性与分层抽样改善不平衡链路质量数据集对评估模型性能的影响，提高评估结果的准确性和稳定性，为给上层协议的优化奠定基础。

High-throughput biomarker segmentation utilizing hierarchical normalized cuts

一种用于获得染色区域分割图像的方法和设备可以包括：量化样本原始图像中生物标志物染色的存在范围，其可以包括基于用户指定的域知识选择数据的域样片；通过对原始图像之内的数据执行频率加权均值平移来对所述原始图像之内的数据进行聚类，以收敛形成分层的多层，每层具有不同的数据分辨率以形成分层数据锥体；分割所述多个均值平移数据图像以确定所述分层数据锥体之内的每个均值平移数据图像中未排除在所述样片之外的数据；在空间上将未排除在所述样片之外的数据映射回所述原始图像以生成最终图像；以及在存储介质上存储所述最终图像以进行进一步分析。

Method and apparatus for sectioning image into plurality of regions

A method and apparatus for sectioning an image into a plurality of regions, by which regions of an image having various features can be extracted stably in a form similar to what humans can perceive, are provided. The method for sectioning an input image into a plurality of regions includes the steps of: converting each pixel of the input image into color coordinates in an arbitrary color space including a luminance (L) axis; forming a plurality of layers and a plurality of bins from the color space; if a distance between the centers of upper last and lower last circles determined using a color, a texture, and a position on an image plane for a pixel is less than a first threshold value, including the upper last and lower last circles in the same cluster and performing the above step on an adjacent lower layer; if the distance between the centers of the upper last and lower last circles is no less than the first threshold value or if the adjacent lower layer does not exist, partitioning the cluster using a last circle including the smallest number of pixels in the color space used in calculating an average color, an average texture, and an average position among the last circles included in the cluster; and performing the above steps on the remaining pixels not included in the upper last or the lower last circle and generating an image graph using clusters obtained for all pixels in the color space.

Image processing method, device, equipment and storage medium

本申请涉及计算机处理技术领域，提供一种图像处理方法、装置、设备和存储介质，主要包括：获取对图像集进行聚类得到的第一连通图；按照所述第一连通图包含的各连接边的介数中心性测度由大到小的顺序，预测所述各连接边是否属于错误连接边；若预测到属于错误连接边的目标连接边，则获取所述第一连通图的重构连通图；所述重构连通图是根据所述第一连通图的边外图重构得到的；所述边外图包括所述第一连通图中所述目标连接边以外的图；根据所述第一连通图和所述重构连通图，确定所述图像集的第二连通图，在节省计算资源和降低时间成本的同时保证图像聚类准确性。

Method, electronic device, and storage medium for providing recommendation service

An electronic device includes a housing, a communication module positioned inside the housing, a processor positioned inside the housing and operatively connected with the communication module, a sensor module operatively connected with the processor, and a memory positioned inside the housing and operatively connected with the communication module, the sensor module, and the processor. The memory stores instructions configured to, when executed, enable the processor to gather data related to a first user, send a request for a user group corresponding to a first category among a plurality of categories to an external server using the communication module, obtain the user group corresponding to the first category based on at least part of the data related to the first user from the external server using the communication module, and provide information about at least one second user in the obtained user group.

Information processing apparatus, method and program

An information processing apparatus, which creates a tree structure used by a recognition apparatus which recognizes specific information using the tree structure, including a memory unit which stores data including the information to be recognized and data not including the information so as to correspond to a label showing whether or not the data includes the information, a recognition device which recognizes the information and outputs a high score value when the data including the information is input, and a grouping unit which performs grouping of the recognition devices using a score distribution obtained when the data is input into the recognition devices.

Model training method, image information management method, image recognition method and device

本发明公开了一种模型训练方法、图像信息管理方法、图像识别方法及装置。所述模型训练方法包括：对初始图像处理模型进行随机初始化；所述初始图像处理模型包括特征提取分支和原型构建分支；固定原型构建分支参数，将训练图像集输入至初始图像处理模型，得到第一输出结果；根据第一输出结果确定损失函数，利用损失函数对特征提取分支的初始化参数进行更新；固定特征提取分支更新后的参数，将训练图像集输入至初始图像处理模型中，得到第二输出结果；根据第二输出结果确定损失函数，利用损失函数对原型构建分支的初始化参数进行更新；根据特征提取分支更新后的参数构建商品指纹提取模型，以使商品指纹提取模型输出的商品指纹具备更好的区分性。

Vehicle-mounted laser point cloud roadmarking recognition methods based on graph structure

本发明涉及一种基于图结构的车载激光点云道路标线识别方法，包括以下步骤：步骤S1：基于强度校正的标线分割和小尺寸目标提取,基于分割结果提取小尺寸标线对象；步骤S2：根据获得的小尺寸标线对象，基于标线的形状特征和邻域空间结构构建由参考标线指向邻域对象的空间与语义关系图;步骤S3：基于空间与语义关系图，自定义图结构节点与边存储编码规则，构建图结构的特征描述子；步骤S4：根据图结构的特征描述子，构建并训练深度神经网络；步骤S5:利用训练后的深度神经网络从车载激光点云目标分割结果中直接提取不同类型的标线。本发明能够实现对复杂城市场景中多类型标线的自动识别与精细分类。

SEMI-SUPERVISED LEARNING PROCESS SYSTEM FOR CLASSIFICATION OF DATA FROM DISCRIMINANT PARAMETERS

Ce procédé d'apprentissage semi-supervisé pour la classification de données comporte : - une étape (E40) de définition d'un graphe génératif dont les noeuds sont constitués par des prototypes représentant ces données, les prototypes étant étiquetés ; - une étape (E40) d'attribution d'un poids à chacun des arcs de ce graphe, ce poids étant estimé par la densité des données autour de cet arc ; - une étape (E50) de propagation desdites étiquettes dans le graphe, jusqu'à un état de convergence ; et - une étape (E60) de classification des données en fonction de l'étiquette des prototypes après ladite propagation (E50). This semi-supervised learning method for the classification of data comprises: a step (E40) of definition of a generative graph whose nodes are constituted by prototypes representing these data, the prototypes being labeled; a step (E40) for assigning a weight to each of the arcs of this graph, this weight being estimated by the density of the data around this arc; a step (E50) for propagating said tags in the graph, to a state of convergence; and a step (E60) of classifying the data according to the label of the prototypes after said propagation (E50).

High-throughput biomarker segmentation utilizing hierarchical normalized cuts

一种用于获得染色区域分割图像的方法和设备可以包括：量化样本原始图像中生物标志物染色的存在范围，其可以包括基于用户指定的域知识选择数据的域样片；通过对原始图像之内的数据执行频率加权均值平移来对所述原始图像之内的数据进行聚类，以收敛形成分层的多层，每层具有不同的数据分辨率以形成分层数据锥体；分割所述多个均值平移数据图像以确定所述分层数据锥体之内的每个均值平移数据图像中未排除在所述样片之外的数据；在空间上将未排除在所述样片之外的数据映射回所述原始图像以生成最终图像；以及在存储介质上存储所述最终图像以进行进一步分析。

Data graph focus mode

A request is received to temporarily identify nodes that are related to one or more selected nodes of a plurality of nodes for a data graph data structure that includes a plurality of relationships between a plurality of nodes arranged in a predetermined spatial arrangement. One or more related nodes that are related to the one or more selected nodes are identified. A display generates only the one or more selected nodes and each of the one or more related nodes and the relationships thereof in an updated spatial arrangement that defines relatively less cumulative space between the one or more selected nodes and the one or more related nodes than the predetermined spatial arrangement. A request to stop generating the updated spatial arrangement is received. The display stops generating the predetermined spatial arrangement in response to receiving the request to stop generating the updated spatial arrangement.