CN114881892B - Remote sensing image characteristic discretization method and device based on II-type fuzzy rough model - Google Patents

Remote sensing image characteristic discretization method and device based on II-type fuzzy rough model Download PDF

Info

Publication number
CN114881892B
CN114881892B CN202210776562.XA CN202210776562A CN114881892B CN 114881892 B CN114881892 B CN 114881892B CN 202210776562 A CN202210776562 A CN 202210776562A CN 114881892 B CN114881892 B CN 114881892B
Authority
CN
China
Prior art keywords
fuzzy
remote sensing
sensing image
membership degree
discretization
Prior art date
Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
Active
Application number
CN202210776562.XA
Other languages
Chinese (zh)
Other versions
CN114881892A (en
Inventor
陈琼
黄小猛
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Tsinghua University
Original Assignee
Tsinghua University
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by Tsinghua University filed Critical Tsinghua University
Priority to CN202210776562.XA priority Critical patent/CN114881892B/en
Priority to PCT/CN2022/105555 priority patent/WO2022258077A2/en
Priority to LU503147A priority patent/LU503147B1/en
Publication of CN114881892A publication Critical patent/CN114881892A/en
Application granted granted Critical
Publication of CN114881892B publication Critical patent/CN114881892B/en
Active legal-status Critical Current
Anticipated expiration legal-status Critical

Links

Images

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T5/00—Image enhancement or restoration
    • G06T5/70—Denoising; Smoothing
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N5/00—Computing arrangements using knowledge-based models
    • G06N5/04—Inference or reasoning models
    • G06N5/048—Fuzzy inferencing
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/12—Computing arrangements based on biological models using genetic models
    • G06N3/126—Evolutionary algorithms, e.g. genetic algorithms or genetic programming
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N7/00—Computing arrangements based on specific mathematical models
    • G06N7/02—Computing arrangements based on specific mathematical models using fuzzy logic
    • G06N7/023—Learning or tuning the parameters of a fuzzy system
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T7/00—Image analysis
    • G06T7/10—Segmentation; Edge detection
    • G06T7/11—Region-based segmentation
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T2207/00—Indexing scheme for image analysis or image enhancement
    • G06T2207/10—Image acquisition modality
    • G06T2207/10032—Satellite or aerial image; Remote sensing

Landscapes

  • Engineering & Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • Theoretical Computer Science (AREA)
  • General Physics & Mathematics (AREA)
  • Software Systems (AREA)
  • Health & Medical Sciences (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Fuzzy Systems (AREA)
  • Biophysics (AREA)
  • Mathematical Physics (AREA)
  • Artificial Intelligence (AREA)
  • Automation & Control Theory (AREA)
  • General Engineering & Computer Science (AREA)
  • Data Mining & Analysis (AREA)
  • Evolutionary Computation (AREA)
  • Computing Systems (AREA)
  • Molecular Biology (AREA)
  • Bioinformatics & Computational Biology (AREA)
  • Computational Linguistics (AREA)
  • Biomedical Technology (AREA)
  • Bioinformatics & Cheminformatics (AREA)
  • General Health & Medical Sciences (AREA)
  • Evolutionary Biology (AREA)
  • Physiology (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • Genetics & Genomics (AREA)
  • Algebra (AREA)
  • Computational Mathematics (AREA)
  • Mathematical Analysis (AREA)
  • Mathematical Optimization (AREA)
  • Pure & Applied Mathematics (AREA)
  • Image Analysis (AREA)
  • Image Processing (AREA)

Abstract

The invention provides a remote sensing image characteristic discretization method and a device based on a II-type fuzzy rough model, wherein the method comprises the following steps: acquiring target remote sensing image data, and extracting mixed pixels from the target remote sensing image data, wherein each mixed pixel respectively comprises spectral response characteristics of multiple surface feature types; determining the main membership degree of each mixed pixel corresponding to each ground feature type according to the mixed pixels; calculating the secondary membership degree of each mixed pixel belonging to each surface feature type according to the primary membership degree; determining a II-type fuzzy rough set of each object type according to the primary membership degree and the secondary membership degree; and carrying out characteristic discretization processing on the target remote sensing image data to obtain an optimal discretization result. According to the embodiment of the invention, the fuzzy component in the remote sensing image characteristic discretization process is described by the primary membership degree and the secondary membership degree corresponding to the mixed pixel, the primary membership degree is used for carrying out the fuzzy discretization process, the primary membership degree is further fuzzified by the secondary membership degree, the uncertainty of the mixed pixel is accurately quantified, and the accurate discretization result is obtained.

Description

Remote sensing image characteristic discretization method and device based on II-type fuzzy rough model
Technical Field
The invention relates to the field of remote sensing image feature extraction, in particular to a remote sensing image feature discretization method and device based on a II-type fuzzy rough model.
Background
Remote sensing has been widely used in various fields of economic and social development as an advanced technical means. The spatial resolution, the time resolution, the spectral resolution and the radiation resolution are gradually improved, so that the remote sensing data have obvious big data characteristics. Due to the diversity, the penetrability and the complexity of the spatial distribution of the ground feature elements, the spectrum signal of each pixel in the remote sensing image records different land coverage types, and the pixels are called mixed pixels. Feature discretization, one of the most influential data preprocessing techniques, plays an important role in the fields of knowledge discovery and data mining, which are widely applied to industrial control. The method can convert continuous features into discrete features which are closer to the representation of a knowledge layer, so that the data is easier to understand, use and interpret, and therefore the efficiency of remote sensing data processing is improved and the method is suitable for learning algorithms which need discrete data as input. In the related art, in order to realize the remote sensing image feature discretization, a feature discretization algorithm of the remote sensing image is usually based on the assumption that one sample only belongs to a single category, and uncertainty caused by a mixed pixel cannot be described. Or in order to simplify the operation of the type II fuzzy set, the secondary membership degree of the mixed image element is defined as a constant. Although the fuzzy rough model quantifies uncertain information by introducing membership of the pixel to each category, the decomposition model of the mixed pixel has larger error, which causes that the distribution information of the data is not uniform, the uncertainty of the data cannot be accurately described, and the data precision is reduced.
Disclosure of Invention
Therefore, the technical problem to be solved by the invention is to overcome the defect that the uncertainty caused by the fact that the characteristic discretization algorithm in the prior art cannot accurately quantize and evaluate the mixed pixel, so that the remote sensing image characteristic discretization method and the remote sensing image characteristic discretization device based on the II-type fuzzy rough model are provided.
According to a first aspect, the embodiment of the invention provides a remote sensing image feature discretization method based on a fuzzy rough model type II, which includes the following steps: acquiring target remote sensing image data, and extracting mixed pixels from the target remote sensing image data, wherein each mixed pixel respectively comprises spectral response characteristics of multiple surface feature types; determining the main membership degree of each mixed pixel corresponding to each ground feature type according to the mixed pixels; calculating the secondary membership degree of each mixed pixel belonging to each ground object type according to the primary membership degree; determining a II-type fuzzy rough set of each ground object type according to the primary membership degree and the secondary membership degree; and carrying out characteristic discretization processing on the target remote sensing image data to obtain an optimal discretization result.
Optionally, determining the main membership degree of each mixed pixel corresponding to each surface feature type according to the mixed pixels, including: iteratively calculating a fuzzy mean vector and a fuzzy covariance matrix of a preset fuzzy partition matrix, wherein the preset fuzzy partition matrix is composed of membership degrees of the mixed pixels corresponding to various ground object types; and determining the main membership degree of each mixed pixel corresponding to each ground feature type according to the fuzzy partition matrix when the iterative computation meets the iterative termination condition.
Optionally, determining the main membership degree of each mixed pixel corresponding to each ground feature type according to a fuzzy partition matrix when iteration calculation meets an iteration termination condition, including: and determining the abundance corresponding to each mixed pixel according to the fuzzy partition matrix, and taking the abundance as the main membership degree of each ground object type corresponding to each mixed pixel.
Optionally, calculating a secondary membership degree of each mixed pixel belonging to each surface feature type according to the primary membership degree, including: determining a hard segmentation matrix according to a fuzzy segmentation matrix when iteration calculation meets an iteration termination condition; determining a set formed by pixels belonging to each surface feature type according to the hard segmentation matrix; calculating upper approximation, lower approximation, a positive domain, a negative domain and a boundary domain of a set in an approximation space; and determining the secondary membership degree of each mixed pixel belonging to each ground object type according to the upper approximation, the lower approximation, the positive domain, the negative domain and the boundary domain.
Optionally, performing feature discretization processing on the target remote sensing image data to obtain an optimal discretization result, including: acquiring an initial breakpoint set of a mixed pixel from remote sensing image data; initializing a target remote sensing image data population based on the number of breakpoints of the initial breakpoint set; iteratively executing a genetic algorithm on individuals of the target remote sensing image data population to determine an optimal discretization result; the discretization scheme corresponding to the initialized target remote sensing image data population is an initial discretization scheme, and each population individual corresponds to one discretization result.
Optionally, iteratively executing a genetic algorithm on the individual of the target remote sensing image data population to determine an optimal discretization result, including: determining a fuzzy relation between the mixed pixels based on Euclidean distances between the mixed pixels; calculating the average approximate precision of the type II fuzzy rough set according to the fuzzy relation; determining the reduction amplitude of the number of the breakpoints corresponding to the individual target remote sensing image data population according to the number of the breakpoints of the initial breakpoint set; determining a fitness function of the type II fuzzy rough set according to the amplitude of the reduction of the number of the fault points and the average approximate precision; and determining the fitness value of the II-type fuzzy rough set according to the fitness function of the II-type fuzzy rough set, and taking the individual of each target remote sensing image data population corresponding to the optimal fitness value as an optimal discretization result.
Optionally, the fitness function of the type II fuzzy rough set is expressed by the following formula:
Figure DEST_PATH_IMAGE001
，
wherein, the first and the second end of the pipe are connected with each other,αandβis a weight coefficient, <' >DL is the magnitude of the reduction in the number of breakpoints,
Figure DEST_PATH_IMAGE002
is the average approximation accuracy.
According to a second aspect, the embodiment of the invention provides a remote sensing image feature discretization device based on a fuzzy rough model type II, which comprises: the mixed pixel extraction unit is configured to acquire target remote sensing image data and extract mixed pixels from the target remote sensing image data, and each mixed pixel comprises spectral response characteristics of multiple surface feature types; the main membership degree determining unit is configured to determine the main membership degree of each mixed pixel corresponding to each ground feature type according to the mixed pixels; the secondary membership degree determining unit is configured to calculate the secondary membership degree of each mixed pixel belonging to each ground feature type according to the primary membership degree; the fuzzy rough set determining unit is configured to determine a II-type fuzzy rough set of each ground object type according to the primary membership degree and the secondary membership degree; and the optimal discretization result determining unit is configured to perform characteristic discretization processing on the target remote sensing image data to obtain an optimal discretization result.
According to a third aspect, an embodiment of the present invention provides a non-transitory computer-readable storage medium, which stores computer instructions, and when the computer instructions are executed by a processor, the method for discretizing a feature of a remote sensing image based on a fuzzy rough model of type II according to any one of the embodiments of the first aspect is implemented.
According to a fourth aspect, embodiments of the present invention provide a computer apparatus comprising at least one processor; and a memory communicatively coupled to the at least one processor; wherein the memory stores instructions executable by the at least one processor, and the instructions are executable by the at least one processor to perform the method for discretizing a feature of a remote sensing image based on a blurred rough model of type II according to any of the embodiments of the first aspect.
The technical scheme of the invention has the following advantages:
the invention provides a remote sensing image characteristic discretization method and a device based on a II-type fuzzy rough model, wherein the method comprises the following steps: acquiring target remote sensing image data, and extracting mixed pixels from the target remote sensing image data, wherein each mixed pixel comprises spectral response characteristics of various surface feature types; determining the main membership degree of each mixed pixel corresponding to each ground feature type according to the mixed pixels; calculating the secondary membership degree of each mixed pixel belonging to each surface feature type according to the primary membership degree; determining a II-type fuzzy rough set of each object type according to the primary membership degree and the secondary membership degree; and carrying out characteristic discretization processing on the target remote sensing image data to obtain an optimal discretization result. The embodiment of the invention combines the rough set and the fuzzy set, describes the fuzzy component in the discretization process of the remote sensing image characteristic by the primary membership degree and the secondary membership degree corresponding to the mixed pixel, fuzzifies the discretization process by the primary membership degree, and further fuzzifies the primary membership degree by the secondary membership degree, thereby accurately quantifying and evaluating the uncertainty of the mixed pixel and obtaining a more accurate discretization result.
Drawings
In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings used in the description of the embodiments or the prior art will be briefly described below, and it is obvious that the drawings in the following description are some embodiments of the present invention, and other drawings can be obtained by those skilled in the art without creative efforts.
Fig. 1 is a flowchart of a specific example of a remote sensing image feature discretization method based on a type II fuzzy rough model in embodiment 1 of the present invention;
fig. 2 is a specific exemplary analysis diagram of a discretization method of remote sensing image features based on a fuzzy rough model type II in embodiment 1 of the present invention;
fig. 3 is a diagram illustrating a structure of a remote sensing image feature discretization apparatus based on a type II fuzzy rough model in embodiment 2 of the present invention;
fig. 4 is a diagram showing an example of the structure of a computer apparatus in embodiment 4 of the present invention.
Detailed Description
The technical solutions of the present invention will be described clearly and completely with reference to the accompanying drawings, and it should be understood that the described embodiments are some, but not all embodiments of the present invention. All other embodiments, which can be obtained by a person skilled in the art without making any creative effort based on the embodiments in the present invention, belong to the protection scope of the present invention.
In the description of the present invention, it should be noted that the terms "center", "upper", "lower", "left", "right", "vertical", "horizontal", "inner", "outer", etc. indicate orientations or positional relationships based on the orientations or positional relationships shown in the drawings, and are only for convenience of description and simplification of description, but do not indicate or imply that the device or element referred to must have a specific orientation, be constructed and operated in a specific orientation, and thus, should not be construed as limiting the present invention. Furthermore, the terms "first," "second," and "third" are used for descriptive purposes only and are not to be construed as indicating or implying relative importance.
In the description of the present invention, the fuzzy-coarse model is a more powerful model of uncertainty data analysis than the fuzzy set and the coarse set. And introducing a fuzzy set on the basis of the rough set, and describing the correlation between the samples by adopting a similarity relation to replace an equivalent relation of the rough set. As the popularization of the fuzzy rough model, the II type fuzzy rough model can provide more accurate uncertainty analysis capability. The fuzzy rough model of II type fuzzifies the membership function values of the fuzzy set again, so that the fuzzy phenomenon can be described more deeply.
In the description of the present invention with respect to a formula,expmeans natural constants in higher mathematicseAn exponential function of the base.infDenotes the infimum bound, which is the largest lower bound of a set.supRepresenting the supremum bound, is the smallest upper bound of a collection.
In addition, the technical features involved in the different embodiments of the present invention described below may be combined with each other as long as they do not conflict with each other.
Example 1
The embodiment provides a discretization method for remote sensing image features based on a II-type fuzzy rough model, which comprises the following steps as shown in fig. 1:
s11: acquiring target remote sensing image data, and extracting mixed pixels from the target remote sensing image data, wherein each mixed pixel respectively comprises spectral response characteristics of multiple surface feature types.
Specifically, the components of the mixed pixel spectral signal are called end members, and each end member corresponds to the spectral response characteristic of one of the terrain types.
S12: and determining the main membership degree of each mixed pixel corresponding to each ground feature type according to the mixed pixels.
Specifically, determining the main membership degree of each corresponding ground object type according to the mixed pixels is to calculate a fuzzy partition matrix through iteration; and under the condition that the iterative computation meets the iterative termination condition, determining the abundance corresponding to each mixed pixel, and taking the abundance as the main membership degree of each ground type corresponding to each mixed pixel. The abundance corresponding to each mixed pixel refers to the abundance of the end members of the mixed pixels.
In practical application, the fuzzy partition matrix is composed of the membership degree of each mixed pixel to the class number of the classification scheme. The fuzzy mean vector and the fuzzy covariance matrix in the fuzzy partition matrix can be represented by the above membership.
S13: and calculating the secondary membership degree of each mixed pixel belonging to each ground feature type according to the primary membership degree.
Specifically, calculating the secondary membership degree of each mixed pixel belonging to each surface feature type means determining the secondary membership degree according to the distribution of the mixed pixels in the boundary area of the rough set. The distribution condition of the mixed pixels in the boundary area of the rough set comprises determining a set formed by pixels belonging to various surface feature types; and calculating the distribution area of the set in the approximate space to determine the secondary membership degree of the mixed pixel belonging to each surface feature type. Wherein, the distribution region of the collection in the approximate space comprises: upper approximation, lower approximation, positive domain, negative domain, boundary domain in the approximation space are aggregated.
In practical applications, the approximation space refers to a coarse approximation space (U,T) WhereinUA set of mixed picture elements is represented,Tthe number of the wave bands of the remote sensing image is represented.
S14: and determining a II-type fuzzy rough set of each ground object type according to the primary membership degree and the secondary membership degree.
Specifically, in the process of determining the primary membership degree of each mixed pixel corresponding to each ground feature type and the secondary membership degree of each mixed pixel belonging to each ground feature type, the fuzzy component in the remote sensing image characteristic discretization process is described by the primary membership degree and the secondary membership degree, the primary membership degree is subjected to fuzzy discretization process, the secondary membership degree is further subjected to fuzzification, and the II type fuzzy rough set of each ground feature type is determined through the determined primary membership degree and the determined secondary membership degree.
S15: and carrying out characteristic discretization processing on the target remote sensing image data to obtain an optimal discretization result.
Specifically, the discretization is to adopt a certain specific method to divide the continuous features into a plurality of subintervals and associate the subintervals with the candidate breakpoints. Therefore, the characteristic discretization processing of the target remote sensing image can be regarded as the selection of the candidate breakpoint. The process of carrying out characteristic discretization processing on the target remote sensing image data to obtain an optimal discretization result refers to iterative selection of candidate breakpoints through a genetic algorithm; determining the fitness function of individuals in the population according to the reduction range of the number of the candidate fault points in each iteration process and the average approximate precision of the type II fuzzy rough set; and evaluating the discretization result according to the determined fitness function of the individuals in the population, and obtaining the optimal discretization result.
In practical application, the initial discretization scheme is determined by acquiring an initial breakpoint set of a mixed pixel in remote sensing image data.
The invention provides a remote sensing image characteristic discretization method based on a II-type fuzzy rough model, which comprises the following steps: acquiring target remote sensing image data, and extracting mixed pixels from the target remote sensing image data, wherein each mixed pixel respectively comprises spectral response characteristics of multiple surface feature types; determining the main membership degree of each mixed pixel corresponding to each ground feature type according to the mixed pixels; calculating the secondary membership degree of each mixed pixel belonging to each surface feature type according to the primary membership degree; determining a II-type fuzzy rough set of each object type according to the primary membership degree and the secondary membership degree; and carrying out characteristic discretization processing on the target remote sensing image data to obtain an optimal discretization result. According to the embodiment of the invention, the rough set and the fuzzy set are combined, the fuzzy component in the remote sensing image characteristic discretization process is described by the primary membership and the secondary membership corresponding to the mixed pixel, the discretization process is fuzzified by the primary membership, and the primary membership is further fuzzified by the secondary membership, so that the uncertainty of the mixed pixel is accurately quantified and evaluated, and a more accurate discretization result is obtained.
In an optional embodiment of the present invention, in the step S12, determining the main membership degree of each mixed pixel corresponding to each surface feature type according to the mixed pixels includes:
(1) Iteratively calculating a fuzzy mean vector and a fuzzy covariance matrix of a preset fuzzy partition matrix, wherein the preset fuzzy partition matrix is composed of membership degrees of the mixed pixels corresponding to various ground object types;
specifically, the preset fuzzy partition matrix can be expressed by the following formula:
Figure DEST_PATH_IMAGE003
，
wherein, the first and the second end of the pipe are connected with each other,F s (X k ) RepresentUTo middlekPixel elementX k For is tosThe degree of membership of a class,s∈{1,2,…,g}，gthe number of the categories is indicated and,k∈{1,2,…,n}，nrepresentUThe number of the middle pixels.
In the practical application of the method, the material is,F s (X k ) Satisfies the following conditions: 0 is less than or equal toF s (X k )≤1，
Figure DEST_PATH_IMAGE004
，q∈{1,2,…,g}。
Specifically, the fuzzy mean vector of the preset fuzzy partition matrix can be expressed by the following formula:
Figure DEST_PATH_IMAGE005
，
wherein the content of the first and second substances,
Figure DEST_PATH_IMAGE006
which represents the fuzzy average of the mean,i∈{1,2,…,m}，mthe number of the wave bands is shown,
Figure DEST_PATH_IMAGE007
，wthe weight is represented by a weight that is,wis more than or equal to 1, and the reaction solution,x ik is shown askThe picture element is at the firstiPixel values over a band.
Specifically, the fuzzy covariance matrix of the preset fuzzy partition matrix can be expressed as follows:
Figure DEST_PATH_IMAGE008
wherein the content of the first and second substances,σ mms the mean of the fuzzy covariance,j∈{1,2,…,m}，
Figure DEST_PATH_IMAGE009
。
specifically, iteratively calculating the fuzzy mean vector and the fuzzy covariance matrix of the preset fuzzy partition matrix includes determining the preset fuzzy partition matrix from the fuzzy mean vector and the fuzzy covariance matrix.
In practical application, the preset fuzzy partition matrix is composed of the membership degree of pixel pairs and categories, the membership degree of the pixel pairs and the categories can be determined by a fuzzy mean vector and a fuzzy covariance matrix, and the membership degree of the pixel pairs and the categories can be expressed according to the following formula:
Figure DEST_PATH_IMAGE010
，
wherein, the first and the second end of the pipe are connected with each other,P`(s) Is as followssThe prior probability of the occurrence of a class,
Figure DEST_PATH_IMAGE011
。
in practical application, forsThe determination of the prior probability of the occurrence of the category belongs to the mature prior art, and is not repeated in this application.
(2) And determining the main membership degree of each mixed pixel corresponding to each ground feature type according to the fuzzy partition matrix when the iterative computation meets the iterative termination condition.
Specifically, the iteration termination condition can be expressed by the following formula:
Figure DEST_PATH_IMAGE012
，
wherein, the first and the second end of the pipe are connected with each other,θfor the number of iteration steps of the algorithm,εis an error threshold.
In an optional embodiment of the present invention, determining the primary membership degree of each mixed pixel corresponding to each surface feature type according to the fuzzy partition matrix when the iterative computation satisfies the iterative termination condition includes: and determining the abundance corresponding to each mixed pixel according to the fuzzy partition matrix, and taking the abundance as the main membership degree of each ground object type corresponding to each mixed pixel.
Specifically, the determination of the main membership degree of each mixed pixel corresponding to each surface feature type is related to the weight and the abundance of each end member in the mixed pixel.
Figure DEST_PATH_IMAGE013
WhereineIs a natural number. The main membership degree of each mixed pixel corresponding to each surface feature type can be expressed according to the following formula:
Figure DEST_PATH_IMAGE014
，
Figure DEST_PATH_IMAGE015
，
wherein, the first and the second end of the pipe are connected with each other,P s (x) Representing mixed pixelsxThe corresponding abundance of the protein is shown in the figure,
Figure DEST_PATH_IMAGE016
the secondary membership degree of each mixed pixel belonging to each ground object type is represented,J x the value range of the main membership degree is represented,uwhich represents a degree of primary membership,
Figure DEST_PATH_IMAGE017
represents the minimum value of the main membership degree of each mixed pixel corresponding to each surface feature type,
Figure DEST_PATH_IMAGE018
and expressing the maximum value of the main membership degree of each mixed pixel corresponding to each ground feature type.
In the practical application of the method, the material is,edetermine (a)wThe number of the middle weights is the same as the weight,ethe greater the value of (a) is,P s (x) The more elements are contained, that is, the larger the value range of the main membership degree is, and meanwhile, a large amount of calculation is also brought. Weight ofwNot only the convexity and concavity of the discretization result, but also the sharing degree of the mixed pixels among various types is controlled. Research results show that when the weight is 2, the clustering effect can be best with a larger probability. If it is noteIf the value of (1) is greater than 1, different weights are introduced, the fuzzy clustering result can be considered more comprehensively, but the clustering quality caused by the introduction of the weights cannot be ensured, so that the discretization result has errors and is unstable. Thus to ensure the stability of the discretization results and reduce the complexity, in an alternative embodiment, one chooseseHas a value of 1 andwthe value is 2. At this time, the process of the present invention,
Figure DEST_PATH_IMAGE019
. It should be understood in relation towAnd witheIncluding but not limited to the embodiments described,wand witheThe value of (a) is only required to be used for ensuring the stability of the discretization result and reducing the complexity.
In practical application, for
Figure DEST_PATH_IMAGE020
The main membership degree of each mixed pixel corresponding to each ground feature type can be expressed according to the following formula:
Figure DEST_PATH_IMAGE021
，
Figure DEST_PATH_IMAGE022
。
in an alternative embodiment of the invention, the fuzzy partition matrix is determined by iteratively computing a fuzzy mean vector and a fuzzy partition matrix. And determining the abundance corresponding to the mixed pixel through the fuzzy partition matrix, and taking the abundance as the main membership degree of each surface feature type corresponding to the mixed pixel. Thereby realizing the fuzzification discretization process with the main membership degree.
In an optional embodiment of the present invention, in step S13, calculating a secondary membership degree of each mixed pixel to each surface feature type according to the primary membership degree includes:
(1) And determining a hard segmentation matrix according to the fuzzy segmentation matrix when the iteration calculation meets the iteration termination condition.
Specifically, the hard partition matrix is determined by modifying the maximum value of each column in the fuzzy partition matrix when the iteration termination condition is satisfied to 1 and modifying the other values in the corresponding columns to 0, thereby completing the determination of the hard partition matrix.
(2) And determining a set formed by the image elements belonging to the types of the ground objects according to the hard partition matrix.
Specifically, the set of image elements belonging to each surface feature type can be expressed by the following formula:
Figure DEST_PATH_IMAGE023
，
wherein Xs represents a set formed by pixels belonging to the s-th terrain type in the mixed pixel set, and Cs (x) is the membership degree of the pixel x in the hard segmentation matrix C to the s-th category.
In practical application, the hard partition matrix determines the corresponding position with the membership value of 1 in each column, and judges the mixed pixel set according to the determined corresponding position with the membership value of 1 in each columnUWhether the mixed image element belongs to each surface feature type or not is achieved, and therefore the set formed by the image elements belonging to each surface feature type is determined according to the hard segmentation matrix.
(3) And calculating upper approximation, lower approximation, positive domain, negative domain and boundary domain of the set in an approximation space.
In particular, a set of pixel elements belonging to each surface feature type is calculatedX s The upper approximation, the lower approximation, the positive domain, the negative domain, and the boundary domain in the approximation space include: determining an equivalence class set of the mixed pixel set; and calculating upper approximation, lower approximation, positive domain, negative domain and boundary domain of the set formed by the image elements belonging to the terrain types in the approximation space according to the determined equivalence class set.
Specifically, the set of equivalence classes of the mixed pel set can be expressed by the following formula:
Figure DEST_PATH_IMAGE024
，
wherein the content of the first and second substances,U|IND(T) An equivalence class set representing a set of mixed picture elements,T={t 1 ,…,t m }，
Figure DEST_PATH_IMAGE025
representing arbitrary picture elementsxAnd any pixelyBelonging to mixed picture elementsX，
Figure DEST_PATH_IMAGE026
Indicates the existence and the pixel of any wave band txPixel elementyRespectively corresponding to the equivalent relation of the wave bands.
In particular, the amount of the solvent to be used,X s the upper approximation in the approximation space can be calculated as follows:
Figure DEST_PATH_IMAGE027
，
wherein the content of the first and second substances,T * (X s ) To representX s An upper approximation in an approximation space.
In particular, the amount of the solvent to be used,X s the following approximation in the approximation space can be calculated as follows:
Figure DEST_PATH_IMAGE028
，
wherein the content of the first and second substances,T * (X s ) Representing a lower approximation of the set in an approximation space.
In particular, the amount of the solvent to be used,X s the positive domain in the approximation space can be calculated as follows:
Figure DEST_PATH_IMAGE029
，
wherein the content of the first and second substances,POS T (X s ) Representing a positive domain assembled in an approximation space.
In particular, the amount of the solvent to be used,X s the negative domain in the approximation space can be calculated as follows:
Figure DEST_PATH_IMAGE030
，
wherein the content of the first and second substances,NGT T (X s ) Representing the negative domain of the set in approximation space.
In particular, the amount of the solvent to be used,X s the boundary domain in the approximation space can be calculated as follows:
Figure DEST_PATH_IMAGE031
，
wherein, the first and the second end of the pipe are connected with each other,BN T (X s ) Representing a boundary domain that is assembled in an approximation space.
(4) And determining the secondary membership degree of each mixed pixel belonging to each ground object type according to the upper approximation, the lower approximation, the positive domain, the negative domain and the boundary domain.
Specifically, determining the secondary membership degree of each mixed pixel belonging to each surface feature type according to the upper approximation, the lower approximation, the positive domain, the negative domain and the boundary domain, and the method comprises the following steps: determining the probability of a set formed by pixels of which the pixels belong to each surface feature type according to the upper approximation, the lower approximation, the positive domain, the negative domain and the boundary domain; and determining the secondary membership degree of each mixed pixel belonging to each surface feature type according to the probability of the determined set formed by the pixels of the pixel belonging to each surface feature type.
Specifically, determining the probability of the set of pixels belonging to each surface feature type according to the upper approximation, the lower approximation, the positive domain, the negative domain and the boundary domain means determining the distribution condition of the pixels according to the upper approximation, the lower approximation, the positive domain, the negative domain and the boundary domain; and determining the probability of the image elements belonging to the set formed by the image elements of all the surface feature types according to the Bayesian theorem under the determined distribution condition.
Is as an example，POS T (X s ) Is attributed to in the mixed pel setX s Is used to form a set of picture elements,NGT T (X s ) Is not attributed to in the mixed pixel setX s Is used to form a set of picture elements,BN T (X s ) Is that in the mixed pixel set it cannot be certain to belong toX s Is used to form a set. Therefore, the temperature of the molten metal is controlled,BN T (X s ) Is the uncertainty field of the mixed picture element. Fig. 2 shows a specific example analysis diagram of a discretization method for remote sensing image features based on a fuzzy rough model type II in the embodiment of the present invention, where each rectangle representsU|IND(T) An equivalence class of (1). The circular area corresponding to reference numeral 22 representsX s The octagonal region corresponding to reference numeral 24 indicatesT * (X s ) And the rectangular region corresponding to reference numeral 23 representsT * (X s ) OrPOS T (X s ). The region other than the octagon corresponding to reference numeral 24 indicatesNGT T (X s ) The region inside the octagon corresponding to reference numeral 24 and outside the rectangle corresponding to reference numeral 23 indicatesBN T (X s ). When picture elementxWhen present in either the positive or negative domain,xand withX s Is determined whenxWhen it is present in the boundary field,xandX s is uncertain. Namely, the distribution situation of the image elements is determined according to the upper approximation, the lower approximation, the positive domain, the negative domain and the boundary domain.
Specifically, with the determined distribution situation, determining the probability of the image element belonging to the set formed by the image elements of all surface feature types according to the Bayesian theorem, including: the probability that a picture element belongs to a set of picture elements of each surface feature type can be expressed by the following formula:
Figure DEST_PATH_IMAGE032
，
wherein, the first and the second end of the pipe are connected with each other,E[i]representing the second on the set of picture elementsiThe number of the equivalence classes is equal to or greater than the number of the equivalence classes,G[s]is shown assThe types of the information to be transmitted are,P(G[s]|E[i]) Representing equivalence classesiMiddle classsThe proportion of the pixels in the image is increased,P(E[i]|G[s]) Is expressed in all categories assBelong to equivalence class in picture elementsiThe proportion of the pixels in the image is increased,P(E[i]) Represents equivalence classesiIn the mixed image element setUThe probability of occurrence of (a) is,P(G[s]) Representing categoriessIn the picture elementUThe probability of occurrence of (c).
Specifically, determining the secondary membership degree of each mixed pixel belonging to each surface feature type according to the probability of the determined set formed by the pixels of the pixel belonging to each surface feature type, comprising the following steps: the subordination degree of each mixed pixel belonging to each surface feature type can be expressed according to the following formula:
Figure DEST_PATH_IMAGE033
。
in an optional embodiment of the invention, an upper approximation, a lower approximation, a positive domain, a negative domain and a boundary domain of a set in an approximation space are calculated by determining a set formed by pixels of all surface feature types; determining the probability of the pixel belonging to a set formed by pixels of all surface feature types according to the upper approximation, the lower approximation, the positive domain, the negative domain and the boundary domain; and determining the secondary membership degree of each mixed pixel belonging to each surface feature type according to the probability of the determined set formed by the pixels of the pixel belonging to each surface feature type. In this process, since the discretization scheme is equivalent to dividing the set of mixed pels into multiple equivalence classes, elements in the same equivalence class have the same attribute value. Then the probability that the mixed pixels in the same equivalence class belong to each class can be considered as the same, and for the mixed pixels in different equivalence classes, the probability that the mixed pixels belong to each class is different, and the uncertainty caused by the difference is further fuzzification of the main membership. Therefore, the process of determining the secondary membership degree is equivalent to further fuzzifying the primary membership degree through the secondary membership degree, so that the uncertainty of the mixed pixel is accurately quantified and evaluated.
In an alternative embodiment of the present invention, in the step S14, the above-mentioned step
Figure DEST_PATH_IMAGE034
The rough set of type II blur for each type of terrain is expressed by the following formula:
Figure DEST_PATH_IMAGE035
。
in an optional embodiment of the present invention, in step S15, performing a feature discretization process on the target remote sensing image data to obtain an optimal discretization result, where the step includes:
(1) Acquiring an initial breakpoint set of a mixed pixel from remote sensing image data;
specifically, acquiring the initial breakpoint set of the mixed pixel from the remote sensing image data refers to acquiring the initial breakpoint set from the input remote sensing image features. The acquisition of the initial breakpoint set belongs to the mature prior art, and is not described herein again.
(2) Initializing a target remote sensing image data population based on the number of breakpoints of the initial breakpoint set;
specifically, the discretization is to adopt a certain specific method to divide the continuous features into a plurality of subintervals and associate the subintervals with the candidate breakpoints. Therefore, the characteristic discretization processing of the target remote sensing image can be regarded as selection of candidate breakpoints, and each discretization scheme corresponds to a division result on the mixed pixel set. The initial target remote sensing image data population refers to the fact that the number of breakpoints of an initial breakpoint set is used as the individual length of the initial population, and therefore initialization of the target remote sensing image data population is completed.
Exemplarily, assuming that the number of initial breakpoints is 10, since each population uses binary coding, the length of each individual is the number of initial breakpoints, that is, the length of each population is 10 bits, and corresponds to 10 breakpoints of the initial breakpoint set respectively.
(3) Iteratively executing a genetic algorithm on individuals of the target remote sensing image data population to determine an optimal discretization result; the discretization scheme corresponding to the initialized target remote sensing image data population is an initial discretization scheme, and each population individual corresponds to one discretization result.
Specifically, the discretization scheme corresponding to the individual in the initial population is an initial discretization scheme, and the discretization result corresponding to the initial discretization scheme is an initial discretization result.
Illustratively, assuming that the number of individuals in the target remote sensing image data population is 50, and the discretization scheme corresponding to the size of the target remote sensing image data population is 50, in each iteration, the 50 individuals in the population undergo selection, variation and intersection, i.e., the evolution function of the genetic algorithm, to update the population, so that the 50 individuals in the population of the next generation are different from the 50 individuals of the previous generation.
In an optional embodiment of the present invention, iteratively executing a genetic algorithm on individuals of the target remote sensing image data population to determine an optimal discretization result includes:
(1) Determining a fuzzy relation between the mixed pixels based on Euclidean distances between the mixed pixels;
specifically, the fuzzy relationship between the mixed pixels can be expressed by the following formula:
Figure DEST_PATH_IMAGE036
，
Figure DEST_PATH_IMAGE037
，
wherein, the first and the second end of the pipe are connected with each other,d(x,y) To representxAndythe distance in degrees of euclidean of (c),x h andy h respectively representxAndyin thathPixel values over a band.
(2) Calculating the average approximate precision of the type II fuzzy rough set according to the fuzzy relation;
in an alternative embodiment of the present invention, calculating the average approximation accuracy of the type II fuzzy rough set based on the fuzzy relation comprises: calculating the upper approximation and the lower approximation of the type II fuzzy rough set according to the fuzzy relation; and calculating the average approximation accuracy of the fuzzy rough set II according to the determined upper approximation and lower approximation.
Specifically, the upper approximation of the type II blur roughness set may be expressed by the following formula:
Figure DEST_PATH_IMAGE038
，
wherein, the first and the second end of the pipe are connected with each other,
Figure DEST_PATH_IMAGE039
，
Figure DEST_PATH_IMAGE040
，u2 is one of the primary degrees of membership,
Figure DEST_PATH_IMAGE041
，J y2 the value range of the corresponding membership degree is represented,a(y2) Representing mixed pixelsyA degree of sub-membership of 2,
Figure DEST_PATH_IMAGE042
and
Figure DEST_PATH_IMAGE043
are respectively
Figure DEST_PATH_IMAGE044
Minimum and maximum primary membership.
Specifically, the following approximation of the type II fuzzy rough set can be expressed as follows:
Figure DEST_PATH_IMAGE045
，
wherein the content of the first and second substances,
Figure DEST_PATH_IMAGE046
，
Figure DEST_PATH_IMAGE047
，u1 is one of the primary degrees of membership,
Figure DEST_PATH_IMAGE048
，J y1 the value range of the corresponding membership degree is represented,a(y1) Representing mixed pixelsyA secondary degree of membership of 1.
Specifically, the average approximation accuracy of the type II fuzzy rough set can be expressed as follows:
Figure DEST_PATH_IMAGE049
，
wherein, the first and the second end of the pipe are connected with each other,
Figure DEST_PATH_IMAGE050
in order to average the accuracy of the approximation,
Figure DEST_PATH_IMAGE051
indicating the approximate accuracy of the type II fuzzy rough set.
Specifically, the approximation accuracy of the type II fuzzy rough set can be expressed as follows:
Figure DEST_PATH_IMAGE052
wherein the content of the first and second substances,x，y1，y2∈U。
illustratively, for
Figure DEST_PATH_IMAGE053
Then, the approximation accuracy of the type II fuzzy rough set can be expressed as follows:
Figure DEST_PATH_IMAGE054
。
(3) Determining the reduction amplitude of the number of the breakpoints corresponding to the individual target remote sensing image data population according to the number of the breakpoints of the initial breakpoint set;
specifically, determining the reduction amplitude of the number of the breakpoints corresponding to the individual target remote sensing image data population according to the number of the breakpoints of the initial breakpoint set comprises the following steps: determining the length of the individual as the number of initial breakpoints; determining the selected length number of the individual as the number of the broken points of the corresponding individual; and determining the difference value between the length of the individual and the number of the individual broken points as the reduction range of the number of the broken points corresponding to the individual of the target remote sensing image data population.
Exemplarily, assuming that the number of initial break points is 10, since each population individual uses binary coding, the length of the individual is the number of initial break points, that is, the length of each population individual is 10 bits, which respectively correspond to 10 break points of the initial break point set. Each bit in the binary code corresponds to a candidate breakpoint, and the values of '1' and '0' represent that the breakpoint is selected and unselected, respectively. For each population individual, determining the number of selected binary codes as the number of breakpoint of the corresponding individual, namely determining the number of binary bits with the value of '1' to obtain the number of breakpoint of the individual. And determining the difference value between the length of the individual and the number of the individual broken points as the reduction amplitude of the number of the broken points of the corresponding individual, namely subtracting the number of the broken points of the individual from 10 to be equal to the reduction amplitude of the number of the broken points of the individual. For example, the binary code of an individual is 1110000111, the number of the broken points corresponding to the individual is 6, and the reduction amplitude of the number of the broken points of the individual is 4.
Specifically, the magnitude of the reduction in the number of breakpoints refers to the number of reductions in the number of breakpoints of individual population in each iteration process. The discrete effect is measured by comparing the quantity of the reduction of the number of the break points of the population in the iterative process, and the larger the reduction of the number of the break points, the better the corresponding discrete effect.
(4) Determining a fitness function of the type II fuzzy rough set according to the amplitude of the reduction of the number of the fault points and the average approximate precision;
specifically, the fitness function of the type II fuzzy rough set is used for calculating the fitness value of each target remote sensing image data population individual, and the fitness function is formed by weighted summation of the amplitude of the reduction of the number of the fault points and the average approximate precision of the type II fuzzy rough set.
In an alternative embodiment of the invention, the fitness function for the type II fuzzy rough set is expressed by the following equation:
Figure DEST_PATH_IMAGE055
，
wherein the content of the first and second substances,αandβis a weight coefficient ofDL is the magnitude of the reduction in the number of breakpoints,
Figure DEST_PATH_IMAGE056
is the average approximation accuracy.
Specifically, the weight coefficient is selected according to an actual working condition, and the reasonability of the weight setting is generally judged according to characteristics of a data set and experimental observation, which is not specifically limited in the present application.
(5) And determining the fitness value of the II-type fuzzy rough set according to the fitness function of the II-type fuzzy rough set, and taking the individual of each target remote sensing image data population corresponding to the optimal fitness value as the optimal discretization result.
Specifically, a plurality of discretization schemes are used as population individuals in a genetic algorithm, the individuals with the maximum fitness value are searched through iterative calculation through the evolution function of the genetic algorithm, and the individuals with the maximum fitness value are used as the individuals with the optimal fitness value. And the discretization scheme corresponding to the optimal fitness value is the optimal discretization scheme.
Illustratively, assuming that there are 50 individuals in the target remote sensing image data population, the fitness function corresponding to the population individuals has 50 values. In each iteration, these 50 individuals in the population will undergo evolution to update the population. So that 50 individuals in the population of the next generation are different from 50 individuals of the previous generation. In each iteration process, the global variable records the individual with the highest fitness value in 50 individuals. When the fitness value of the next generation of existing individuals is higher than the fitness value of the individual of the global variable record, the global variable is updated with the individual having the higher fitness value. After all iterations are finished, the global variable records the optimal individual, and the discretization scheme corresponding to the optimal individual is the optimal discretization scheme.
In an alternative embodiment of the invention, the fitness function for the type II fuzzy rough set is constructed by calculating the average approximation accuracy of the type II fuzzy rough set and determining the magnitude of the reduction in the number of breakpoints. And determining an individual of an optimal fitness value through a genetic algorithm, describing fuzzy components in the remote sensing image characteristic discretization process by using the primary membership and the secondary membership corresponding to the mixed pixels, fuzzifying the discretization process by using the primary membership, further fuzzifying the primary membership by using the secondary membership, and obtaining a more accurate discretization result through a constructed fitness function of a II-type fuzzy rough set.
Example 2
This embodiment provides a remote sensing image characteristic discretization device based on a type II fuzzy rough model, as shown in fig. 3, fig. 3 is a connection diagram of the remote sensing image characteristic discretization device based on the type II fuzzy rough model according to an optional embodiment of the present invention, including: the method comprises a mixed pixel extracting unit 31, a main membership degree determining unit 32, a secondary membership degree determining unit 33, a fuzzy rough set determining unit 34 and an optimal discretization result determining unit 35.
The mixed pixel extracting unit 31 is configured to acquire target remote sensing image data and extract mixed pixels from the target remote sensing image data, wherein each mixed pixel respectively comprises spectral response characteristics of multiple surface feature types. For details, reference may be made to the related description of step S11 in any of the above method embodiments, and details are not repeated herein.
And the main membership determining unit 32 is configured to determine the main membership of each mixed pixel corresponding to each surface feature type according to the mixed pixels. For details, reference may be made to the related description of step S12 in any of the above method embodiments, and details are not repeated herein.
And a secondary membership degree determining unit 33 configured to calculate a secondary membership degree of each mixed pixel belonging to each surface feature type according to the primary membership degree. For details, reference may be made to the related description of step S13 in any of the above method embodiments, and details are not repeated herein.
And a fuzzy rough set determining unit 34 configured to determine a type II fuzzy rough set for each feature type according to the primary and secondary membership degrees. For details, reference may be made to the related description of step S14 in any of the above method embodiments, and details are not repeated herein.
And the optimal discretization result determining unit 35 is configured to perform feature discretization on the target remote sensing image data to obtain an optimal discretization result. For details, reference may be made to the description related to step S15 of any of the above method embodiments, and details are not repeated here.
The invention provides a remote sensing image characteristic discretization device based on a II-type fuzzy rough model, which comprises: and the mixed pixel extraction unit 31 is configured to acquire target remote sensing image data and extract mixed pixels from the target remote sensing image data, wherein each mixed pixel respectively comprises spectral response characteristics of multiple surface feature types. And the main membership determining unit 32 is configured to determine the main membership of each mixed pixel corresponding to each surface feature type according to the mixed pixels. And the secondary membership degree determining unit 33 is configured to calculate the secondary membership degree of each mixed pixel belonging to each surface feature type according to the primary membership degree. And a fuzzy rough set determining unit 34 configured to determine type II fuzzy rough sets for each feature type according to the primary and secondary membership degrees. And the optimal discretization result determining unit 35 is configured to perform characteristic discretization processing on the target remote sensing image data to obtain an optimal discretization result. According to the embodiment of the invention, the rough set and the fuzzy set are combined, the fuzzy component in the remote sensing image characteristic discretization process is described by the primary membership and the secondary membership corresponding to the mixed pixel, the discretization process is fuzzified by the primary membership, and the primary membership is further fuzzified by the secondary membership, so that the uncertainty of the mixed pixel is accurately quantified and evaluated, and a more accurate discretization result is obtained.
In an alternative embodiment of the present invention, the primary membership determining unit 32 includes: and iteratively calculating the subunits and determining the subunits according to the main membership degree. The details can be seen in the related description of determining the main membership degree of each mixed image element corresponding to each surface feature type according to the mixed image elements in any of the above method embodiments.
And the iterative computation subunit is configured to iteratively compute a fuzzy mean vector and a fuzzy covariance matrix of a preset fuzzy partition matrix, wherein the preset fuzzy partition matrix consists of membership degrees of the mixed pixels corresponding to the types of the ground objects.
And the main membership degree determining subunit is configured to determine the main membership degree of each mixed pixel corresponding to each ground object type according to the fuzzy partition matrix when the iterative computation meets the iterative termination condition.
In an optional embodiment of the present invention, the sub-membership determining subunit includes an abundance determining subunit configured to determine, according to the fuzzy partition matrix, an abundance corresponding to each mixed pixel, and use the abundance as a main membership of each ground type corresponding to each mixed pixel. For details, reference may be made to the description of determining the main membership degree of each mixed pixel corresponding to each surface feature type according to the fuzzy partition matrix when the iterative computation satisfies the iterative termination condition in any of the above method embodiments.
In an alternative embodiment of the present invention, the secondary membership degree determining unit 33 includes: the hard segmentation matrix determines a subunit, the pixel set determines a subunit, the approximate space boundary determines a subunit, and the secondary membership determines a subunit. The details can be seen in the related description of calculating the secondary membership degree of each mixed image element to each surface feature type according to the primary membership degree in any of the above method embodiments.
A hard segmentation matrix determination subunit configured to determine a hard segmentation matrix from the fuzzy segmentation matrix when the iteration computation satisfies the iteration termination condition.
And the image element set determining subunit is configured to determine a set formed by image elements belonging to each surface feature type according to the hard segmentation matrix.
An approximation space boundary determining subunit configured to calculate an upper approximation, a lower approximation, a positive domain, a negative domain, a boundary domain aggregated in an approximation space.
And the secondary membership determining subunit is configured to determine the secondary membership of each mixed pixel belonging to each surface feature type according to the upper approximation, the lower approximation, the positive domain, the negative domain and the boundary domain.
In an optional embodiment of the present invention, the optimal discretization result determining unit 35 includes an initial breakpoint set obtaining subunit, a population initializing subunit, and an optimal discretization result determining subunit. For details, reference may be made to the relevant description of discretization processing on target remote sensing image data to obtain an optimal discretization result in any of the above method embodiments.
And the initial breakpoint set acquisition subunit is configured to acquire an initial breakpoint set of the mixed pixel from the remote sensing image data.
And the population initialization subunit is configured to initialize the target remote sensing image data population based on the number of breakpoints of the initial breakpoint set. The discretization scheme corresponding to the initialized target remote sensing image data population is an initial discretization scheme, and each population individual corresponds to one discretization result.
And the optimal discretization result determining subunit is configured to iteratively execute a genetic algorithm on the individuals of the target remote sensing image data population to determine an optimal discretization result.
In an optional embodiment of the present invention, the optimal discretization result determining subunit includes: the method comprises a pixel-to-pixel fuzzy relation determining subunit, a calculating subunit, an amplitude determining subunit with reduced number of broken points, a fitness function determining subunit and a fitness value determining subunit. For details, reference may be made to the description related to the determination of the optimal discretization result in any of the above method embodiments, regarding iteratively executing the genetic algorithm on the individual of the target remote sensing image data population.
And the inter-pixel fuzzy relation determining subunit is configured to determine the fuzzy relation among the mixed pixels based on Euclidean distances among the mixed pixels.
A calculating subunit configured to calculate an average approximation accuracy of the type II fuzzy rough set according to the fuzzy relation.
And the amplitude determining subunit is configured to determine the amplitude of the reduction of the number of the breakpoints corresponding to the target remote sensing image data population individuals according to the number of the breakpoints of the initial breakpoint set.
And the fitness function determining subunit is configured to determine a fitness function of the type II fuzzy rough set according to the reduction amplitude of the number of the fault points and the average approximate precision.
And the fitness value determining subunit is configured to determine the fitness value of the type II fuzzy rough set according to the fitness function of the type II fuzzy rough set, and take the individual of each target remote sensing image data population corresponding to the optimal fitness value as an optimal discretization result.
The invention provides a remote sensing image characteristic discretization device based on a II-type fuzzy rough model, which comprises: and a mixed pixel extracting unit 31 configured to acquire target remote sensing image data and extract mixed pixels from the target remote sensing image data, wherein each mixed pixel respectively comprises spectral response characteristics of multiple surface feature types. And the main membership determining unit 32 is configured to determine the main membership of each mixed pixel corresponding to each surface feature type according to the mixed pixels. And a secondary membership degree determining unit 33 configured to calculate a secondary membership degree of each mixed pixel belonging to each surface feature type according to the primary membership degree. And a fuzzy rough set determining unit 34 configured to determine a type II fuzzy rough set for each feature type according to the primary and secondary membership degrees. And the optimal discretization result determining unit 35 is configured to perform feature discretization on the target remote sensing image data to obtain an optimal discretization result. According to the embodiment of the invention, the rough set and the fuzzy set are combined, the fuzzy component in the remote sensing image characteristic discretization process is described by the primary membership and the secondary membership corresponding to the mixed pixel, the discretization process is fuzzified by the primary membership, and the primary membership is further fuzzified by the secondary membership, so that the uncertainty of the mixed pixel is accurately quantified and evaluated, and a more accurate discretization result is obtained.
Example 3
An embodiment of the present invention also provides a non-transitory computer storage medium having stored thereon computer-executable instructions that can perform the method described in any of the method embodiments above. The storage medium may be a magnetic Disk, an optical Disk, a Read-Only Memory (ROM), a Random Access Memory (RAM), a Flash Memory (Flash Memory), a Hard Disk (Hard Disk Drive, abbreviated as HDD), a Solid State Drive (SSD), or the like; the storage medium may also comprise a combination of memories of the kind described above.
Example 4
An embodiment of the present invention further provides a computer device, as shown in fig. 4, fig. 4 is a schematic structural diagram of a computer device according to an alternative embodiment of the present invention, and the computer device may include at least one processor 41, at least one communication interface 42, at least one communication bus 43, and at least one memory 44, where the communication interface 42 may include a Display (Display) and a Keyboard (Keyboard), and the alternative communication interface 42 may also include a standard wired interface and a wireless interface. The Memory 44 may be a Random Access Memory (RAM) or a non-volatile Memory (non-volatile Memory), such as at least one disk Memory. The memory 44 may alternatively be at least one memory device located remotely from the aforementioned processor 41. Wherein processor 41 may be associated with the apparatus described in fig. 3, an application program is stored in memory 44, and processor 41 calls program code stored in memory 44 for performing the steps of the method described in any of the method embodiments above.
The communication bus 43 may be a Peripheral Component Interconnect (PCI) bus, an Extended Industry Standard Architecture (EISA) bus, or the like. The communication bus 43 may be divided into an address bus, a data bus, a control bus, and the like. For ease of illustration, only one thick line is shown in FIG. 4, but that does not indicate only one bus or one type of bus.
The memory 44 may include a volatile memory (RAM), such as a random-access memory (RAM); the memory may also include a non-volatile memory (english: flash memory), such as a Hard Disk Drive (HDD) or a solid-state drive (SSD); the memory 44 may also comprise a combination of the above-mentioned kinds of memories.
The processor 41 may be a Central Processing Unit (CPU), a Network Processor (NP), or a combination of CPU and NP.
The processor 41 may further include a hardware chip. The hardware chip may be an application-specific integrated circuit (ASIC), a Programmable Logic Device (PLD), or a combination thereof. The PLD may be a Complex Programmable Logic Device (CPLD), a field-programmable gate array (FPGA), a General Array Logic (GAL), or any combination thereof.
Optionally, the memory 44 is also used to store program instructions. Processor 41 may call program instructions to implement the methods described in any of the embodiments of the present invention.
It should be understood that the above examples are only for clarity of illustration and are not intended to limit the embodiments. Other variations and modifications will be apparent to persons skilled in the art in light of the above description. And are neither required nor exhaustive of all embodiments. And obvious variations or modifications derived therefrom are intended to be within the scope of the invention.

Claims (9)

1. A remote sensing image feature discretization method based on a II-type fuzzy rough model is characterized by comprising the following steps:
acquiring target remote sensing image data, and extracting mixed pixels from the target remote sensing image data, wherein each mixed pixel respectively comprises spectral response characteristics of multiple surface feature types;
determining the main membership degree of each mixed pixel corresponding to each ground feature type according to the mixed pixels;
calculating the secondary membership degree of each mixed pixel belonging to each ground feature type according to the primary membership degree;
determining a II-type fuzzy rough set of each ground object type according to the primary membership degree and the secondary membership degree;
carrying out characteristic discretization processing on the target remote sensing image data to obtain an optimal discretization result;
wherein, the calculating the secondary membership degree of each mixed pixel belonging to each ground feature type according to the primary membership degree comprises:
determining a hard segmentation matrix according to a fuzzy segmentation matrix when iteration calculation meets an iteration termination condition;
determining a set formed by pixels belonging to various surface feature types according to the hard segmentation matrix;
calculating upper approximation, lower approximation, positive domain, negative domain, boundary domain of the set in an approximation space;
and determining the secondary membership degree of each mixed pixel belonging to each ground object type according to the upper approximation, the lower approximation, the positive domain, the negative domain and the boundary domain.
2. The remote sensing image characteristic discretization method based on the II-type fuzzy rough model, according to the claim 1, wherein the determining the main membership degree of each mixed pixel corresponding to each surface feature type according to the mixed pixels comprises:
iteratively calculating a fuzzy mean vector and a fuzzy covariance matrix of a preset fuzzy partition matrix, wherein the preset fuzzy partition matrix is composed of membership degrees of mixed pixels corresponding to various ground object types;
and determining the main membership degree of each mixed pixel corresponding to each ground feature type according to the fuzzy partition matrix when the iterative computation meets the iterative termination condition.
3. The remote sensing image characteristic discretization method based on the II-type fuzzy rough model as claimed in claim 2, wherein the determining the main membership degree of each mixed pixel corresponding to each surface feature type according to the fuzzy partition matrix when the iterative computation meets the iteration termination condition comprises:
and determining the abundance corresponding to each mixed pixel according to the fuzzy partition matrix, and taking the abundance as the main membership degree of each ground object type corresponding to each mixed pixel.
4. The remote sensing image feature discretization method based on the fuzzy rough model type II according to claim 1, wherein the performing the feature discretization on the target remote sensing image data to obtain an optimal discretization result comprises:
acquiring an initial breakpoint set of the mixed pixel from the remote sensing image data;
initializing the target remote sensing image data population based on the number of breakpoints of the initial breakpoint set;
iteratively executing a genetic algorithm on the individual of the target remote sensing image data population to determine an optimal discretization result;
the discretization scheme corresponding to the initialized target remote sensing image data population is an initial discretization scheme, and each population individual corresponds to one discretization result.
5. The remote sensing image characteristic discretization method based on the fuzzy rough model type II according to claim 4, wherein the step of iteratively executing a genetic algorithm on the individuals of the target remote sensing image data population to determine an optimal discretization result comprises the following steps:
determining a fuzzy relation among the mixed pixels based on Euclidean distances among the mixed pixels;
calculating the average approximate precision of the type II fuzzy rough set according to the fuzzy relation;
determining the reduction amplitude of the number of the breakpoints corresponding to the target remote sensing image data population individuals according to the number of the breakpoints of the initial breakpoint set;
determining a fitness function of the II-type fuzzy rough set according to the amplitude of the reduction of the number of the fault points and the average approximate precision;
and determining the fitness value of the II-type fuzzy rough set according to the fitness function of the II-type fuzzy rough set, and taking the individual of each target remote sensing image data population corresponding to the optimal fitness value as the optimal discretization result.
6. The discretization method for remote sensing image characteristics based on type II fuzzy rough model according to claim 5, wherein the fitness function of type II fuzzy rough set is expressed by the following formula:
Figure 976982DEST_PATH_IMAGE001
wherein, the first and the second end of the pipe are connected with each other,αandβas a weight coefficient, | D | is a magnitude that decreases by the number of breakpoints,
Figure 943670DEST_PATH_IMAGE002
is the average approximation accuracy.
7. The utility model provides a remote sensing image characteristic discretization device based on rough model of II type fuzzy, its characterized in that includes:
the mixed pixel extraction unit is configured to acquire target remote sensing image data and extract mixed pixels from the target remote sensing image data, and each mixed pixel comprises spectral response characteristics of multiple surface feature types;
the main membership degree determining unit is configured to determine the main membership degree of each mixed pixel corresponding to each ground feature type according to the mixed pixels;
the secondary membership degree determining unit is configured to calculate the secondary membership degree of each mixed pixel belonging to each ground feature type according to the primary membership degree; wherein, the calculating the secondary membership degree of each mixed pixel belonging to each ground feature type according to the primary membership degree comprises: determining a hard segmentation matrix according to a fuzzy segmentation matrix when iteration calculation meets an iteration termination condition; determining a set formed by pixels belonging to each surface feature type according to the hard segmentation matrix; calculating upper approximation, lower approximation, a positive domain, a negative domain, and a boundary domain of the set in an approximation space; determining the secondary membership degree of each mixed pixel belonging to each ground object type according to the upper approximation, the lower approximation, the positive domain, the negative domain and the boundary domain;
a fuzzy rough set determining unit configured to determine a type II fuzzy rough set of each ground object type according to the primary membership and the secondary membership;
and the optimal discretization result determining unit is configured to perform characteristic discretization processing on the target remote sensing image data to obtain an optimal discretization result.
8. A non-transitory computer-readable storage medium storing computer instructions which, when executed by a processor, implement the method for discretizing features of remote sensing images based on fuzzy coarse model type II as claimed in any one of claims 1 to 6.
9. A computer device, comprising: at least one processor; and a memory communicatively coupled to the at least one processor; wherein the memory stores instructions executable by the at least one processor to perform the method of discretizing a feature of a remotely sensed image based on a fuzzy coarse model of type II as claimed in any one of claims 1-6.
CN202210776562.XA 2022-07-04 2022-07-04 Remote sensing image characteristic discretization method and device based on II-type fuzzy rough model Active CN114881892B (en)

Priority Applications (3)

Application Number Priority Date Filing Date Title
CN202210776562.XA CN114881892B (en) 2022-07-04 2022-07-04 Remote sensing image characteristic discretization method and device based on II-type fuzzy rough model
PCT/CN2022/105555 WO2022258077A2 (en) 2022-07-04 2022-07-13 Remote sensing image feature discretization method and apparatus based on ii-type fuzzy rough model, storage medium, and computer device.
LU503147A LU503147B1 (en) 2022-07-04 2022-07-13 Remote sensing image feature discretization method and apparatus based on type-ii fuzzy rough model, storage medium and computer device

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202210776562.XA CN114881892B (en) 2022-07-04 2022-07-04 Remote sensing image characteristic discretization method and device based on II-type fuzzy rough model

Publications (2)

Publication Number Publication Date
CN114881892A CN114881892A (en) 2022-08-09
CN114881892B true CN114881892B (en) 2022-10-14

Family

ID=82683265

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202210776562.XA Active CN114881892B (en) 2022-07-04 2022-07-04 Remote sensing image characteristic discretization method and device based on II-type fuzzy rough model

Country Status (3)

Country Link
CN (1) CN114881892B (en)
LU (1) LU503147B1 (en)
WO (1) WO2022258077A2 (en)

Families Citing this family (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN117111585B (en) * 2023-09-08 2024-02-09 广东工业大学 Numerical control machine tool health state prediction method based on tolerance sub-relation rough set

Citations (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN106127784A (en) * 2016-07-01 2016-11-16 辽宁工程技术大学 A kind of high-resolution remote sensing image dividing method
AU2020100179A4 (en) * 2020-02-04 2020-03-19 Huang, Shuying DR Optimization Details-Based Injection Model for Remote Sensing Image Fusion
CN111428627A (en) * 2020-03-23 2020-07-17 西北大学 Mountain landform remote sensing extraction method and system
CN112580483A (en) * 2020-12-15 2021-03-30 海南大学 Remote sensing image characteristic discretization method based on rough fuzzy model

Family Cites Families (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN112116610B (en) * 2020-08-24 2024-02-27 中国科学院深圳先进技术研究院 Remote sensing image segmentation method, system, terminal and storage medium

Patent Citations (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN106127784A (en) * 2016-07-01 2016-11-16 辽宁工程技术大学 A kind of high-resolution remote sensing image dividing method
AU2020100179A4 (en) * 2020-02-04 2020-03-19 Huang, Shuying DR Optimization Details-Based Injection Model for Remote Sensing Image Fusion
CN111428627A (en) * 2020-03-23 2020-07-17 西北大学 Mountain landform remote sensing extraction method and system
CN112580483A (en) * 2020-12-15 2021-03-30 海南大学 Remote sensing image characteristic discretization method based on rough fuzzy model

Non-Patent Citations (1)

* Cited by examiner, † Cited by third party
Title
融入空间关系的二型模糊模型高分辨率遥感影像分割;王春艳等;《遥感学报》;20160125(第01期);全文 *

Also Published As

Publication number Publication date
WO2022258077A2 (en) 2022-12-15
LU503147B1 (en) 2023-04-07
CN114881892A (en) 2022-08-09
WO2022258077A3 (en) 2023-09-21

Similar Documents

Publication Publication Date Title
Angelopoulos et al. Image-to-image regression with distribution-free uncertainty quantification and applications in imaging
Sameen et al. Classification of very high resolution aerial photos using spectral‐spatial convolutional neural networks
CN109472199B (en) Image fusion classification method and device
Li et al. DANCE-NET: Density-aware convolution networks with context encoding for airborne LiDAR point cloud classification
Ke et al. Adaptive change detection with significance test
CN113360701B (en) Sketch processing method and system based on knowledge distillation
CN111950643B (en) Image classification model training method, image classification method and corresponding device
Su et al. Deep convolutional neural network–based pixel-wise landslide inventory mapping
US20230186100A1 (en) Neural Network Model for Image Segmentation
CN110197716B (en) Medical image processing method and device and computer readable storage medium
CN110532413B (en) Information retrieval method and device based on picture matching and computer equipment
WO2022179083A1 (en) Image detection method and apparatus, and device, medium and program
CN114881892B (en) Remote sensing image characteristic discretization method and device based on II-type fuzzy rough model
CN116524369B (en) Remote sensing image segmentation model construction method and device and remote sensing image interpretation method
EP3660750B1 (en) Method and system for classification of data
CN112287965A (en) Image quality detection model training method and device and computer equipment
KR20220107940A (en) Method for measuring lesion of medical image
CN114463587A (en) Abnormal data detection method, device, equipment and storage medium
CN113469167A (en) Method, device, equipment and storage medium for recognizing meter reading
CN110533050B (en) Picture geographic information acquisition method and device, computer equipment and storage medium
CN113011376B (en) Marine ship remote sensing classification method and device, computer equipment and storage medium
CN112883898A (en) Ground feature classification method and device based on SAR (synthetic aperture radar) image
Wang et al. A new stochastic simulation algorithm for image-based classification: feature-space indicator simulation
CN116740410B (en) Bimodal target detection model construction method, bimodal target detection model detection method and computer equipment
Bar et al. Facilitating high‐dimensional transparent classification via empirical Bayes variable selection

Legal Events

Date Code Title Description
PB01 Publication
PB01 Publication
SE01 Entry into force of request for substantive examination
SE01 Entry into force of request for substantive examination
GR01 Patent grant
GR01 Patent grant