Plant identification method based on cloud model and TOPSIS (Technique for Order Preference by Similarity to Ideal Solution) method

Technical Field

The invention belongs to the plant identification Technical Field, in particular relates to a method for and cloud model-based TOPSIS plant identification method.

Background Art

Agricultural database in the agricultural information resources in the development of play an important role by, agricultural organizations have at home and building a large batch of agricultural database, wherein, the most famous International in agricultural database system AGRIS  , IFIS, AGRI-COLA, database CABI four major agriculture, agriculture, agriculture organization of the United nations the system database (AGRIS), International food information database (IFIS), agricultural on-line access to the database of the United States Department of agriculture (AGRI-COLA) and International agribio center database (CAB1). My representative plant database with: the National agricultural science (www.agridata.cn) data sharing center, China (http://www.cvh.org.cn) digital plant well, the Chinese plant database (www.plant.csdb.cn), China (www.chinaplant.org) plant web, China (bd.brim.ac.cn) biological diversity information system, such as Image library (www.plantphoto.cn) plant in China.

These database includes abundant professional knowledge, storing a large amount of agricultural related scientific and technical information, as a result of these database professional is strong, and can only be conducted using the key word search query, the operator increased the difficulty of the operation of the database, the use efficiency of the database is reduced, causing a waste of resources. The main is shown as : (1) high requirements on the operator. The use of these database operator must be skilled in the use of computer, database retrieval interface understanding, master search strategy. At the same time, it is necessary to the theme, keyword, mechanism, as a whole, like the same, the search concepts and retrieval means to grasp and understand. However, plant database comprises a user not only a relatively strong field of professional knowledge to experts and agricultural technicians, also includes peasants and non-agricultural science and technology, it is difficult to these most accurate according to the requirements of the input search word ; (2) key word search by the searching mode must be strictly according to the specified format input, only when the assembly is completely matched search results can be obtained only when, this label literally consistent with the retrieved by the retrieval mode, it is very difficult to realize and conceptual the content of the retrieved search results meeting the needs of users, will lead to recall ratio and the precision of retrieval results is relatively low.

To the question above, some researchers use of digital Image processing technology to realize plant identification, according to the authentication result of inquiry of the database, so as to obtain the relevant expertise.

However, at present, the technologies also exist the following problems : (1) the current Image acquisition basic through the scanner, the background is simple, Image segmentation and described is relatively simple, it is difficult to deal with complex background ; (2) the Image database of different researchers are all the same, it is very difficult to compare the performance of the identification effect is good and bad, and is difficult to be used in the practice of ; (3) the current existing Image classification system usually can only process a few plant to hundred , range is relatively small, it is difficult to realize large-scale, multi-class of plant differentiation. However, there have been some researchers proposed:

The literature reports, its provides: "the uncertainty of the knowledge representation in" (the text author was li deyi,   published in 2000 published in "Chinese engineering science" subsection 2 roll section 10 a view to section 73-79 page). The article discloses cloud model that is used for natural language value a certain qualitative concept and its quantitative uncertainty between a representation of a conversion model, comprising:

(1) the definition of the cloud

U is a is accurate numerical representation of the quantitative universe, U C the qualitative concept is on, if the quantitative   ∈ U, and C qualitative concept of a stochastic realization, C indeed fixed degrees to ∈ [O, 1] is a stable tendency to random number of: : U → [O, 1], , . the The distribution of the domain known as the cloud model, referred to as cloud, each Referred to as a kernel.

(2) the digital signature is the cloud

The digital signature is that of the cloud model  , Entropy And ultra entropyTo characterize, reflects the characteristic of the whole C qualitative concept.

Cloud is the desired spatial distribution of the domain, is the most representative of qualitative concept, reflects the cloud center of gravitycloud drop group of this concept. The uncertainty of the qualitative measure of the concept, the concept of joint decision ambiguity and randomness, discloses ambiguity and random referability. entropy In the concept of a qualitative measure of randomness, can be reflected in the kernel on behalf of this qualitative concept of discrete degree of; on the other hand, they are qualitative concept of the measurement of the ambiguity, reflects the concept of domain can be in the space of the range of values of the kernel. for Such a digital characteristic to simultaneously reflect ambiguity and randomness, also reflects the correlation between the two. A measure of the uncertainty of is the entropy , reflects the space wavenumber domain on behalf of the concept of the degree of uncertainty allth aggregation, agglomeration of the kernel. Indirectly expresses the size of the discrete cloud extent and thickness, entropy randomness and ambiguity common decision.

Cloud of three digital is characterized in that the value of the kernel by the tens of thousands of which comprise the entire cloud to, the qualitative expressed in the language of the ambiguity value and randomness is fully integrated together. Because different specific realizing method, a cloud of different types, such as normal cloud model, trapezoidal cloud model, such as half cloud model. Wherein, normal cloud model and trapezoidal cloud model.

There are also books reports, its title is "  Making   Decision   Attribute Multiple:   Application Methods   and" (the book's author is:C.L.Hwang and Yoon   K.,   Berlin Springe press 1981 Publication year). Disclosed in the book according to a limited number of evaluation object and idealised of the proximity of the target degree-discrete approximation method of sequencing an ideal de-sorting method (Technique   for   Order   Preference   by   Similarity   to   Solution   Ideal, abbrebytion TOPSIS). Its principle is based on the original data matrix after normalizing, find limited scheme in the optimal scheme and most noninferiority programme form a space, a certain object to be the evaluation of the space can be regarded as a point on the, according to the the point can be obtained with the optimal scheme and the distance between the most noninferiority programme (commonly used Euclidean distance, also called Euclidean distance), so as to obtain the object and the optimal scheme of the degree which is relatively close to, and can be carried out before the evaluation of the programme.

Content of the invention

In view of the above prior art problems and shortcomings of the, the technical problem to be solved by the invention is to provide a method, and cloud model-based TOPSIS plant identification method, the method can be conveniently, and quickly from the same or of a plurality of plants monoids from the saved in the database of a large number of plant sample retrieval of the measured plant, the identification of the plant.

In order to solve the above-mentioned problem, the invention adopts the following technical scheme:

A TOPSIS method and cloud model-based plant identification method, which is characterized in that the method first of all has constructed the plant the appearance characteristics of a sample database; then utilize the trapezoidal cloud model the appearance of the plant to be tested is the characteristics of appearance of the plant sample database for comparison, obtained with the plant to be tested is compared with the database of the membership grade of the specimen, the preliminary identification of the plant to be measured; when the authentication result is a plurality of time, re-use of normal cloud model to calculate the exact match search result, obtained with the plant to be tested is compared with the parameter of the sample database; TOPSIS finally utilizing the comprehensive evaluation of the membership, identifying a plant, the specific steps are as follows:

(1), construct the plant the appearance characteristics of a sample database;

(2), using trapezoidal cloud model, calculating the membership grade of the plant to be measured, preliminary differentiation of the measured plant;

(3), to judge whether the membership grade of the plant to be tested is less than or equal to 1, if the membership grade of the plant to be tested is less than or equal to 1, then the transfer to the step (5), if the membership grade of the plant to be tested is less than or equal to 1 is equal to 1, then the transfer to the step (4);

(4), using normal cloud model, for the membership grade 1 of the plant to be measured of the sample and the normal cloud model calculation, to obtain the accurate computational plant to be tested;

(5), using TOPSIS method, overall evaluation of the membership grade of the measured to identify the plant.

The above-mentioned step (1) in the construct plant of the appearance characteristics of a sample database, the sample can be used for the database to identify plant types of plant appearance feature, the appearance of the characteristic value for the BCM software.

The above-mentioned step (2) the use of the trapezoidal cloud model, calculating the membership grade of the plant to be measured, preliminary differentiation of the measured plant, the operation steps are as follows:

(21), determining a desired cloud model section of trapezoid   : According to plant appearance characteristic of the appearance characteristics of a sample library in the value range, determine a desired cloud model section of trapezoid , Wherein To this appearance is characterized in the lower limit value, For the appearance of the upper limit value of the characteristic value;

(22), computing scalaris cloud model entropy : By using trapezoidal of the cloud model yunhe drop cloud , the appearance of the plant to expand the range of values of the characteristic, the expansion is set to the interval [-3 , + 3 ];

(23), using trapezoidal of the cloud model to the measured data, the appearance of the described plant is: through the ladder cloud expectations and entropy determination scalaris cloud desired curve equation:

Wherein, X point cloud trapezoidal membership, The desired lower limit of the interval, The section to the desired upper limit, Is a trapezoid cloud entropy ;

(24), using trapezoidal cloud model, calculating the membership grade of the plant to be measured: the plant to be tested with all shapes of appearance in the specimen base respectively, using trapezoidal cloud model characteristic value for contrast analysis, of the tested plants of each of the appearance characteristics .

The above-mentioned step (4) utilization of the normal cloud model in, for the membership grade 1 of the plant to be measured of the sample and the normal cloud model calculation, to obtain the accurate membership grade of the measured plant, the operation steps are as follows:

(41), to determine the expected value of the model according to: according to plant the appearance characteristics of a sample library contour characteristic value, is determined according to the expected value of the model , The intermediate section value appearance characteristics;

(42), entropy of the model is calculated according to , Entropy The formula of: ;

(43), can be determined through normal cloud desired with entropy according to a desired curve equation, the curve equation as:

Wherein, X point cloud trapezoidal membership, To a desired value, Is a trapezoid cloud entropy ;

(44), to be measured all appearance characteristics of the plant with the step (2) calculate the parameter of the plant to be tested 1 contrast analysis of the specimen, the tested plants of each of the appearance characteristics .

The above-mentioned step (5) the use of French TOPSIS in the, overall evaluation of the membership grade of the measured to identify the plant, its operating steps are as follows:

(51), the evaluation of set membership grade F comprehensive evaluation matrix: if steps (2) to calculate the values of only one or no a is equal to 1 at the time, the evaluation matrix F (2) the resulting membership matrix; otherwise, the evaluation matrix F (3) the results obtained by the calculation;

(52), determining the ideal point evaluation matrix F, the formulas of the ideal point:

Wherein, The ideal point set, In subsection i section of the programme by evaluation of the value of the item j, The largest is the number of the objective function, Is set as a function of the most small targets;

(53), the evaluation matrix F determining the most handicapper, its most handicapper of the formula:

Wherein, As the most almost collection , In subsection i section of the programme by evaluation of the value of the item j, The largest is the number of the objective function, Is set as a function of the most small targets;

(54), in the calculation and evaluation matrix F to the evaluation of the programme is the distance of the ideal point, its formula is:

,

Wherein, As to the evaluation of the programme i is the distance of the ideal point, In subsection i section of the programme by evaluation of the value of the item j, For ideal point the value of the section of a j, n is the number of evaluation of the programme;

(55), in the calculation and evaluation matrix F the evaluation scheme to the most handicapper distance, its formula is:

,

Wherein, I to the evaluation scheme to the most handicapper distance, In subsection i section of the programme by evaluation of the value of the item j, As the most handicapper section j of the value, n is the number of evaluation of the programme;

(56), in the calculation and evaluation matrix F programme being evaluated the relative proximity to the ideal point, its formula is:

,

Wherein, I evaluation is to programme the relative proximity to the ideal point, I to the evaluation scheme to the most handicapper distance, I to the evaluation scheme to the most ideal point distance of the, n is the number of evaluation of the programme;

(57), according to the evaluation matrix F in the evaluation of the programme is the ideal point relatively close to degrees Make good and bad ordering to the case.

The above-mentioned step (22) the calculating scalaris cloud model entropy Its expansion in the range of appearance characteristics of the lower limit value 20%, the trapezoidal cloud model entropy For formulas: entropy =( × 0.2)/ 3.

A of this invention and the cloud model-based method for TOPSIS plant identification method compared with the prior art, has the following effects:

(1), the method using plant of the unknown plant additionality proper authentication, the user effectively avoid the plant database keyword search, high requirements of professional knowledge on the operator, and the search result is not matched with the demand, the search range expansion, improves retrieval precision;

(2), the method utilizes the plant cloud model the appearance of the digitized characteristic information described, the appearance of the plant characteristic information of the uncertainty of the qualitative and quantitative between conversion, at the same time, respectively, using trapezoidal cloud model and normal cloud model of the plant characteristic to different Figures to describe appearance characteristics, in that under the condition of identifying precision, the search range expansion, improves the authentication effect;

(3), the method adopts the final authentication result TOPSIS comprehensive evaluation, can comprehensively, reasonable, accurately to a plurality of evaluation indexes of the sorted or, clear of the evaluation process, objective evaluation results.

Description of drawings

Fig. 1 is a of this invention and the cloud model-based method for method for identifying TOPSIS the flow chart of the plant;

Figure 2 is a schematic diagram of the trapezoidal cloud model;

Figure 3 is a schematic diagram of normal cloud model;

Figure 4 the appearance characteristics of a sample database for bamboo ;

Figure 5 is a trapezoid cloud model expected value sector table (saves the name the bamboo plants);

Figure 6 entropy value table for appearance characteristics;

Figure 7 ladder table cloud model is based on the appearance characteristic database and the specimen of bamboo corresponding membership;

Figure 8 for normal cloud model the expected value Table;

Fig. 9 is a normal cloud model entropy Table;

Figure 10 is based on normal cloud model of the database and the specimen of bamboo corresponding to the membership grade of the appearance.

Mode of execution

The Figures and in particular with the embodiment of the invention for further detailed description.

The embodiment of the plant -bamboo as unknown for distinction embodiment.

With reference to Figure 1, the invention is based on the law and cloud model TOPSIS plant identification method, the steps are as follows:

(1), construct the plant the appearance characteristics of a sample database

The sample can be used for the database to identify plant types of plant appearance feature, the appearance of the characteristic value for the BCM software, selected the stalk are high bamboo, diameter, internode long, stalk wall thickness , such as and blade width integrase structure 6 the numerical parameter is a characteristic parameter; and then from the "network world" selects a 15 kind of common bamboo , the establishment of bamboo the appearance characteristics of a sample database. , See Figure 4.

(2), using trapezoidal cloud model, calculating the membership grade of the plant to be tested, measured plant of the initial authentication, the operation steps are as follows:

(21), determining a desired cloud model section of trapezoid : According to the bamboo the appearance characteristics of a sample database table 1 data, determine an expected value in the section , For example, a value of interval desired interval [1-5], which is expressed as, = 1, = 5  , a result of the conversion as shown in Figure 5.

(22) computing scalaris cloud model entropy

By using trapezoidal of the cloud model yunhe drop cloud , the appearance of the plant to expand the range of values of the characteristic, the expansion is set to the interval [-3 , + 3 ], For appearance characteristics for the spreading range of the lower limit value 20%, so the trapezoidal cloud model entropy Formulas: entropy =( × 0.2)/ 3. Calculate all respectively the appearance characteristics of a plant to be tested the enthropy , The result as shown in Figure 6.

(23), using trapezoidal of the cloud model to the measured data, the appearance of the described plant is: through the ladder cloud expectations and entropy determination scalaris cloud desired curve equation:

Wherein, X point cloud trapezoidal membership;

The desired lower limit of the interval;

The section to the desired upper limit;

Is a trapezoid cloud entropy.

For example, bamboo the appearance characteristics of a specimen in the database a 1st described cloud model the trapezoid, bamboo for the stalk are high 1-5m, appearance characteristic of the same, by using trapezoid cloud model for the digital signature is described in:

The trapezoid of the cloud model with reference to the schematic diagram of Figure 2, Figure 2 in the X-axis X said the stalk are highfor bamboo , the unit is m; Y axis Expressed as the membership grade ; Wherein, The desired lower limit of the interval, The section to the desired upper limit, Is a trapezoid cloud entropy.

(24) use of trapezoid cloud model, calculating the membership grade of the measured plant: based on the computed expected value in the section with entropy, Using trapezoidal cloud description of the digital signature of the plant to be tested, are respectively calculate the bamboo the membership grade of the appearance characteristics, as shown in Figure   7 is shown.

(3), to judge whether the membership grade of the plant to be tested is less than or equal to 1, if the membership grade of the plant to be tested is less than or equal to 1, then the transfer to the step (5), if the membership grade of the plant to be tested is less than or equal to 1 is equal to 1, then the transfer to the step (4);

(4), using normal cloud model, for the membership grade 1 of the plant to be measured of the sample and the normal cloud model calculation, step (2) of the bamboo is obtained 8, 11 and 14 of the three kinds of bamboo contrast to the membership grade 1, to the bamboo using normal cloud model with 8, 11 and 14 are identified, the specific steps are as follows:

(41) is determined according to the expected value of the model

According to the plant the appearance characteristics of a sample library contour characteristic value, is determined according to the expected value of the model , The interval value of the appearance characteristics of the intermediate value, the expected value of the normal cloud model, as shown in Figure 8.

(42) according to the model is the entropy , Its formula is: , According to the [-3 , + 3 ] Between the elements of 99.73%, its entropyThe result of calculation as shown in Figure 9.

(43), can be determined through normal cloud desired with entropy according to a desired curve equation, the curve equation as:

Wherein, X point cloud trapezoidal membership; To a desired value; Is a trapezoid cloud entropy ;

Such as bamboo for the stalk are high 1-5m, appearance characteristic of the same, by using trapezoid cloud the described digital is characterized in:

According to the schematic diagram of the model can be with reference to Figure 3, Figure 3 in the X-axis X said the stalk are highfor bamboo , the unit is m; Y axis Expressed as the membership grade ; Wherein, To a desired value, Is a trapezoid cloud entropy.

(44), to be measured all appearance characteristics of the plant with the steps (1) to calculate the membership grade of the plant to be tested 1 contrast analysis of the specimen, the tested plants of each of the appearance characteristics .

Based on the calculated expected value with entropy, Using trapezoidal cloud described digital characteristic, respectively calculate the bamboo the membership grade of the appearance characteristics, the result as shown in Figure 10.

(5), using TOPSIS method, identifying overall evaluation of the membership grade of the measured plant, the specific steps are as follows:

(51), the evaluation of set membership matrix F comprehensive evaluation

If step (2) of a calculated parameter, 8, 11, 14 are three value 1, so that the evaluation matrix F (3) the results of calculation, that is,

(52), determining the ideal point evaluation matrix F, the formulas of the ideal point:

Wherein, the ideal point set, In subsection i section of the programme by evaluation of the value of the item j, is the J the largest number of the objective function, Is a function of the set of small target, the result of calculation, in order to:

(53), the evaluation matrix F determining the most handicapper, its most handicapper of the formula:

Wherein, As the most almost collection , In subsection i section of the programme by evaluation of the value of the item j, is the J the largest number of the objective function, Is set as a function of the most small targets, if the computation thereof:

(54), in the calculation and evaluation matrix F to the evaluation of the programme is the distance of the ideal point, its formula is:

,

Wherein, As to the evaluation of the programme i is the distance of the ideal point, To j i section of a the value of the item, For ideal point the value of the section of a j, n is the number of evaluation of the programme, its result of calculation is:

(55), in the calculation and evaluation matrix F the evaluation scheme to the most handicapper distance, its formula is:

,

Wherein, I to the evaluation scheme to the most handicapper distance, To j i section of a the value of the item, As the most handicapper section j of the value, the number of the evaluation of the programme to be n, the result of calculation, in order to:

(56), in the calculation and evaluation matrix F programme being evaluated the relative proximity to the ideal point, its formula is:

,

Wherein, I evaluation is to programme the relative proximity to the ideal point; I to the evaluation scheme to the most handicapper distance, I to the evaluation scheme to the most ideal point distance of the, n is the number of evaluation of the programme, its result of calculation is:

The relative proximity of the Is a is between the 0 and 1 the value of between, the "ideal point" Speaking, its relative proximity To 1, the "most handicapper" Speaking, its relative proximity To 0  , therefore, relatively close degrees The large, the programmes close to the ideal point vietnam , is better; on the contrary, relatively close degrees The small, the programme vietnam close to the maximum handicapper, the poor.

(57), according to the evaluation matrix F in the evaluation of the programme is the ideal point relatively close to degrees Make good and bad ordering to the case.

Step (56) the relative proximity Failure, the most likely is bamboo that the bamboo 14 (  brachycladum Schizostachyum), followed by the bamboo 8 (  farinosus Dendrocalamus) and bamboo 11 (  Gigantochloa   atroviolacea  ), authentication completed.

The method of the invention is not limited to Mode of execution of the in the embodiment, according to the technical personnel in the field the technical proposal of this invention the other embodiments, also belongs to the scope of technical innovation of this invention.