CN111460943A - Remote sensing image ground object classification method and system - Google Patents

Remote sensing image ground object classification method and system Download PDF

Info

Publication number
CN111460943A
CN111460943A CN202010212440.9A CN202010212440A CN111460943A CN 111460943 A CN111460943 A CN 111460943A CN 202010212440 A CN202010212440 A CN 202010212440A CN 111460943 A CN111460943 A CN 111460943A
Authority
CN
China
Prior art keywords
remote sensing
sensing image
image data
cluster
clustering
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.)
Pending
Application number
CN202010212440.9A
Other languages
Chinese (zh)
Inventor
杜航原
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.)
Shanxi University
Original Assignee
Shanxi 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 Shanxi University filed Critical Shanxi University
Priority to CN202010212440.9A priority Critical patent/CN111460943A/en
Publication of CN111460943A publication Critical patent/CN111460943A/en
Pending legal-status Critical Current

Links

Images

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
    • G06V20/00—Scenes; Scene-specific elements
    • G06V20/10—Terrestrial scenes
    • G06V20/13—Satellite images
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F18/00—Pattern recognition
    • G06F18/20—Analysing
    • G06F18/23—Clustering techniques
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F18/00—Pattern recognition
    • G06F18/20—Analysing
    • G06F18/24—Classification techniques

Abstract

The invention provides a method and a system for classifying ground objects of remote sensing images, which can improve the accuracy and the robustness of ground object classification results. The method comprises the following steps: s1, acquiring the remote sensing image to be processed, and preprocessing the remote sensing image to be processed to obtain a characteristic vector set of the remote sensing image to be processed; s2, calculating the condition information entropy of the feature vector set about the remote sensing image data to describe the uncertainty of the remote sensing image data in the feature space; s3, iteratively updating the cluster division of the remote sensing image data based on the obtained feature vector set about the conditional information entropy of the remote sensing image data until the number of the classes of the new cluster division is consistent with the number to be classified of the remote sensing image, and executing S4; and S4, outputting the final clustering division as a ground feature classification result. The invention relates to the technical field of remote sensing image processing.

Description

Remote sensing image ground object classification method and system
Technical Field
The invention relates to the technical field of remote sensing image processing, in particular to a method and a system for classifying ground objects of remote sensing images.
Background
The remote sensing technology is a modern comprehensive technology which receives electromagnetic wave information from various ground features on the earth surface from high altitude or outer space, and carries out scanning, photographing, transmission and processing on the information so as to carry out remote control measurement and identification on various ground features and phenomena on the earth surface. The remote sensing technology has the characteristics of wide sensing range, large amount of information, quick information acquisition, short updating period and the like, can help people to continuously know the driving force of the nature and describe the operation process of various phenomena occurring on the planet where the people are located on the one hand, and helps people to obtain the key information required by a decision support system on the other hand, so that policy making and management decision making are more reasonable. The principle of remote sensing imaging is to record and image by detecting the reflection or radiation wave of the earth surface through a sensor on a high-altitude platform, and the remote sensing imaging can record the reflection of various electromagnetic waves such as visible light, invisible light, radiation wave, radar wave and the like. The images obtained by using the remote sensing equipment are only one part of the remote sensing technology, and the analysis and the processing of the images to obtain interesting information are another important task of the remote sensing technology.
The processing of remote sensing images is very complex and generally comprises: image selection, surface feature reflection spectrum analysis, observation data input, image feature and landform analysis, image preprocessing, image statistical feature analysis, key region processing, image enhancement, classification processing, region processing, composite processing, result output and the like. In many links of remote sensing image processing, classification processing is a process of judging and identifying information such as category attribute spatial distribution characteristics of an interested target according to the difference of the interested target on a remote sensing image. People can automatically classify and process the remote sensing images through a computer to achieve the aim of identifying ground features. In practical applications, the classification process is often used to search for ground feature composition, find special features, and the like. Two types of machine learning methods are often used in remote sensing image classification processing:
one type is supervised classification with an artificial neural network as a representative, the application of the supervised classification in remote sensing image processing is very mature, and generally higher classification accuracy can be achieved, but the supervised classification requires to obtain a training subset of data to be processed and correct class attribution thereof;
the other type is unsupervised classification represented by cluster analysis, the unsupervised classification is based on the principle that similar ground objects have similar spectral characteristics, the characteristic values reflecting the ground objects are classified according to similarity and probability theory, and finally the classification meaning is determined by comparing with field data, so that the method has the advantages of no need of any prior knowledge, difficulty in being influenced by human factors and the like.
In the prior art, clustering analysis is the most commonly used unsupervised classification method, and a k-means clustering (k-means) algorithm and an iterative self-organizing data analysis algorithm (ISODATA) are the most used clustering algorithms in remote sensing image processing. Remote sensing image clustering is greatly different from general data clustering, and due to the complexity of natural environment and the complexity of interaction between the natural environment and a remote sensing spectrum, uncertainty always exists in surface feature information extracted from spectral signals recorded by a sensor, which causes that stable and reliable classification results are difficult to obtain in a remote sensing image classification task by using a traditional clustering method (for example, k-means, ISODATA).
Disclosure of Invention
The invention aims to provide a method and a system for classifying ground objects of remote sensing images, which aim to solve the problem that stable and reliable classification results are difficult to obtain in a remote sensing image classification task in the prior art.
In order to solve the above technical problem, an embodiment of the present invention provides a method for classifying ground features of remote sensing images, including:
s1, acquiring the remote sensing image to be processed, and preprocessing the remote sensing image to be processed to obtain a characteristic vector set of the remote sensing image to be processed;
s2, calculating the condition information entropy of the feature vector set about the remote sensing image data to describe the uncertainty of the remote sensing image data in the feature space;
s3, performing iteration S31-S33, and clustering and dividing the remote sensing image data:
s31, determining the quality coefficient of cluster division in the current iteration by using the uncertainty of the cluster division in the feature space in the current iteration and the obtained conditional information entropy of the feature vector set about the remote sensing image data;
s32, according to the obtained quality coefficients, respectively measuring uncertainty of a set S consisting of the remote sensing image data and any two categories in the current iterative clustering division in a symbol space, determining information gain of the set S in the symbol space, combining the two categories in the set S with the minimum information gain into one category, and generating new clustering division;
s33, judging whether the number of the new cluster classification categories is consistent with the number to be classified of the remote sensing images, if so, executing S4, otherwise, returning to S31 to perform the next iteration;
and S4, outputting the final clustering division as a ground feature classification result.
Further, the obtaining the remote sensing image to be processed and preprocessing the remote sensing image to be processed to obtain the feature vector set of the remote sensing image to be processed includes:
obtaining a remote sensing image to be processed, segmenting an interest area according to a ground object classification task, and extracting spectral data of a target spectral band combination in the remote sensing image;
extracting the characteristic vector of the remote sensing image to be processed according to the obtained interest area and the extracted spectral data and forming a set
Figure BDA0002423282950000031
Where M is the number of extracted feature vectors, fjRepresenting the jth feature vector, the feature vectors in set F including: spectral characteristics, texture characteristics, average reflection intensity of ground objects in the remote sensing image expressed by a gray mean value, and dispersion degree of the gray value of each pixel of the remote sensing image expressed by a gray median value and the sum of the average values;
carrying out image enhancement processing based on linear stretching on the remote sensing image data of the interest area to obtain the remote sensing image data
Figure BDA0002423282950000032
Wherein x isiThe image data acquisition method comprises the steps of representing the ith pixel in remote sensing image data, wherein N represents the number of pixels in the image data;
using sets of Gaussian kernel function pairs
Figure BDA0002423282950000033
Performing feature mapping to obtain a mapped feature vector set
Figure BDA0002423282950000034
And to aggregate the feature vectors
Figure BDA0002423282950000035
The feature vectors in (1) obey gaussian distribution; wherein the gaussian kernel function is represented as:
Figure BDA0002423282950000036
wherein κ represents a Gaussian kernel function, xiAnd xjRespectively represent any two pixels in the remote sensing image data X,
Figure BDA0002423282950000037
respectively representing picture elements xi、xjFeature vector collection
Figure BDA0002423282950000038
The value of (a) is selected from,
Figure BDA0002423282950000039
representing a pixel xiAnd xjFeature vector collection
Figure BDA00024232829500000310
Correlation of the above values, γ represents | | | xi(F)-xj(F)||2Standard deviation of (1), xi(F) Representing a pixel xiValues, x, over a set of characteristic vectors Fi(fj) Representing a pixel xiThe value on the jth eigenvector.
Further, the condition information entropy of the feature vector set about the remote sensing image data is expressed as:
Figure BDA0002423282950000041
wherein the content of the first and second substances,
Figure BDA0002423282950000042
representing a set of feature vectors
Figure BDA0002423282950000043
Entropy of condition information about the remote sensing image data X;
Figure BDA0002423282950000044
representing a set of feature vectors in remote-sensing image data XCombination of Chinese herbs
Figure BDA0002423282950000045
The variance of (a) above (b),
Figure BDA0002423282950000046
the calculation formula of (2) is as follows:
Figure BDA0002423282950000047
wherein, muXRepresenting remote-sensed image data X in
Figure BDA0002423282950000048
The method of (a) is not particularly limited,
Figure BDA0002423282950000049
the calculation formula of (2) is as follows:
Figure BDA00024232829500000410
wherein x isa、xbAnd xcRepresenting X in remote-sensing image data XiAny three other picture elements.
Further, before iteratively performing S31-S33, the method further comprises:
carrying out initialization clustering division on remote sensing image data: each picture element is divided into a separate class, i.e.
Figure BDA00024232829500000411
Wherein, C(0)It is indicated that the cluster partitioning is initialized,
Figure BDA00024232829500000412
representing the nth category in the initial cluster partition.
Further, the calculating the quality coefficient of cluster partitioning in the current iteration by using the uncertainty of cluster partitioning in the feature space in the current iteration and the obtained conditional information entropy of the feature vector set about the remote sensing image data includes:
if the current iteration is the m-th iteration, calculating a characteristic vector set
Figure BDA00024232829500000413
Clustering division C for remote sensing image data X in current iteration(m)Conditional information entropy of
Figure BDA00024232829500000414
To describe the uncertainty of the cluster partition in the feature space; wherein the content of the first and second substances,
Figure BDA00024232829500000415
expressed as:
Figure BDA00024232829500000416
wherein the content of the first and second substances,
Figure BDA00024232829500000417
is represented by C(m)Feature vector collection
Figure BDA00024232829500000418
The variance of (a) above (b),
Figure BDA00024232829500000419
the calculation formula of (2) is as follows:
Figure BDA00024232829500000420
wherein x iseIs any pixel in the remote sensing image data X,
Figure BDA00024232829500000421
is C(m)Feature vector collection
Figure BDA00024232829500000422
In the above-mentioned manner, the expectation is that,
Figure BDA00024232829500000423
the calculation formula of (2) is as follows:
Figure BDA0002423282950000051
wherein x isf，xgAnd xhFor X in remote-sensing image data XeAny other 3 pixels;
according to the obtained characteristic vector set
Figure BDA0002423282950000052
Entropy of conditional information on remote-sensing image data X
Figure BDA0002423282950000053
And feature vector sets
Figure BDA0002423282950000054
Partition C with respect to clustering(m)Conditional information entropy of
Figure BDA0002423282950000055
Determining information gain of cluster partition of remote sensing image data X in characteristic space in current iteration
Figure BDA0002423282950000056
Wherein the content of the first and second substances,
Figure BDA0002423282950000057
expressed as:
Figure BDA0002423282950000058
and determining the quality coefficient of cluster division in the current iteration according to the information gain of the cluster division of the remote sensing image data X in the current iteration in the characteristic space.
Further, the clustering division C of the remote sensing image data X generated by the mth iteration(m)Mass coefficient of (omega)(m)Expressed as:
Figure BDA0002423282950000059
wherein, C(t)And (3) representing the clustering division of the remote sensing image data X generated by the t-th iteration, wherein t is more than or equal to 1 and less than or equal to m.
Further, according to the obtained quality coefficient, determining the information gain of the set S in the symbol space by respectively measuring the uncertainty of the set S composed of the remote sensing image data and any two categories in the current iterative cluster division in the symbol space, merging the two categories in the set S with the minimum information gain into one category, and generating a new cluster division, including:
clustering division of the remote sensing image data in the current iteration is expressed as a clustering vector in a symbol space, and a clustering vector set is obtained;
according to the obtained quality coefficient and the cluster vector set, respectively measuring the uncertainty of a set S consisting of any two categories in the remote sensing image data and the current iterative clustering division in a symbol space, and determining the information gain of the set S in the symbol space;
and combining two categories in the set S with the minimum information gain into one category to generate a new cluster partition.
Further, the dividing and representing the clusters of the remote sensing image data in the current iteration as cluster vectors in a symbol space to obtain a cluster vector set includes:
if the current iteration is the mth iteration, the cluster division of the remote sensing image data X generated by the mth iteration is expressed as a set formed by all the classified categories and recorded as
Figure BDA0002423282950000061
Wherein, C(m)For cluster partitioning in the mth iteration,
Figure BDA0002423282950000062
is represented by C(m)Of the kth class, KmIs represented by C(m)The number of categories in (1);
generating the remote sensing image data of the m-th iteration in the symbol spaceClustering partition C of X(m)Expressed as a cluster vector
Figure BDA0002423282950000063
Wherein lm,kRepresenting a clustering vector lmThe k component of (1)<k<Km；
Recording a set formed by a series of clustering vectors generated in the symbol space by the 1 st to m iterations as a set
Figure BDA0002423282950000064
Wherein ltAnd representing the clustering vectors which are generated by the t iteration and correspond to the clustering division of the remote sensing image data X.
Further, the determining, according to the obtained quality coefficient and cluster vector set, the uncertainty of the set S in the symbol space, which is composed of the remote sensing image data and any two categories in the current iterative cluster division, by measuring the remote sensing image data and the uncertainty of the set S in the symbol space, includes:
if the current iteration is the m-th iteration, determining a cluster vector set in the m-th iteration
Figure BDA0002423282950000065
Wherein each cluster vector is associated with an entropy of conditional information of the remote sensing image data X to describe an uncertainty of the remote sensing image data X in each cluster vector, wherein for t ≦ 1 ≦ m,
Figure BDA0002423282950000066
middle t clustering vector ltConditional information entropy H (l) of remote sensing image data Xt| X) is expressed as:
Figure BDA0002423282950000067
wherein, KtIs represented by C(t)Number of classes in, lt,kAs a cluster vector ltThe k component of (a), P (l)t,k| X) represents lt,kConditional probability, P (l), for remote-sensed image data Xt,k| X) is calculated as:
Figure BDA0002423282950000068
wherein x isi(lt) Representing a pixel xiIn the clustering vector ltThe value of (a), i.e. pixel x in the t-th iterationiThe category into which it is divided;
determining a set of clustering vectors in an mth iteration
Figure BDA0002423282950000069
Entropy of conditional information on remote-sensing image data X
Figure BDA00024232829500000610
To describe the uncertainty in the symbol space of the remotely sensed image data X, wherein,
Figure BDA00024232829500000611
expressed as:
Figure BDA0002423282950000071
wherein, ω is(t)Representing the cluster partition C produced by the t-th iteration(t)The mass coefficient of (a);
partition C in clustering(m)Optionally two of
Figure BDA0002423282950000072
And
Figure BDA0002423282950000073
1≤p≠q≤Kmcomputing a set of clustering vectors
Figure BDA0002423282950000074
About a set for each cluster vector
Figure BDA0002423282950000075
To describe the set S of these two classes at eachUncertainty on the cluster vector, wherein,
Figure BDA0002423282950000076
t th clustering vector ltAbout
Figure BDA0002423282950000077
Conditional information entropy of
Figure BDA0002423282950000078
Expressed as:
Figure BDA0002423282950000079
wherein the content of the first and second substances,
Figure BDA00024232829500000710
representing a clustering vector ltWith respect to the set
Figure BDA00024232829500000711
The conditional probability of (a) of (b),
Figure BDA00024232829500000712
expressed as:
Figure BDA00024232829500000713
determining a set of clustering vectors
Figure BDA00024232829500000714
About collections
Figure BDA00024232829500000715
Conditional information entropy of
Figure BDA00024232829500000716
To describe a collection
Figure BDA00024232829500000717
Uncertainty in symbol space, whichIn (1),
Figure BDA00024232829500000718
expressed as:
Figure BDA00024232829500000719
determining a set
Figure BDA00024232829500000720
Information gain in symbol space
Figure BDA00024232829500000721
Figure BDA00024232829500000722
Expressed as:
Figure BDA00024232829500000723
the embodiment of the invention also provides a remote sensing image ground feature classification system, which comprises:
the processing module is used for acquiring the remote sensing image to be processed and processing the remote sensing image to be processed to obtain a characteristic vector set of the remote sensing image to be processed;
the determining module is used for determining the conditional information entropy of the feature vector set about the remote sensing image data so as to describe the uncertainty of the remote sensing image data in the feature space;
the partitioning module is used for iteratively executing S31-S33 and clustering and partitioning the remote sensing image data:
s31, determining the quality coefficient of cluster division in the current iteration by using the uncertainty of the cluster division in the feature space in the current iteration and the obtained conditional information entropy of the feature vector set about the remote sensing image data;
s32, according to the obtained quality coefficients, respectively measuring uncertainty of a set S consisting of the remote sensing image data and any two categories in the current iterative clustering division in a symbol space, determining information gain of the set S in the symbol space, combining the two categories in the set S with the minimum information gain into one category, and generating new clustering division;
s33, judging whether the number of the new cluster classification categories is consistent with the number to be classified of the remote sensing images, if so, executing an output module, otherwise, returning to S31 for next iteration;
and the output module is used for outputting the final clustering division as a ground feature classification result.
The technical scheme of the invention has the following beneficial effects:
in the scheme, the uncertainty in the remote sensing image data is measured by the conditional information entropy, the uncertainty of information between the ground feature information and the spectral data in the remote sensing image generation process can be effectively reflected, the consistency of the description of the ground feature classification information of the remote sensing image is evaluated in the feature space and the symbol space through iterative computation, the cluster division of the ground feature target of the remote sensing image is realized, and the ground feature classification result has higher classification precision and robustness.
Drawings
Fig. 1 is a schematic flow chart of a remote sensing image surface feature classification method according to an embodiment of the present invention;
fig. 2 is a detailed flowchart of a method for classifying a surface feature of a remote sensing image according to an embodiment of the present invention;
fig. 3 is a schematic structural diagram of a remote sensing image surface feature classification system according to an embodiment of the present invention.
Detailed Description
In order to make the technical problems, technical solutions and advantages of the present invention more apparent, the following detailed description is given with reference to the accompanying drawings and specific embodiments.
The invention provides a remote sensing image ground object classification method and system aiming at the problem that a stable and reliable classification result is difficult to obtain in the existing remote sensing image classification task.
Example one
As shown in fig. 1, the method for classifying the ground features of the remote sensing image according to the embodiment of the present invention includes:
s1, acquiring the remote sensing image to be processed, and preprocessing the remote sensing image to be processed to obtain a characteristic vector set of the remote sensing image to be processed;
s2, calculating the condition information entropy of the feature vector set about the remote sensing image data to describe the uncertainty of the remote sensing image data in the feature space;
s3, performing iteration S31-S33, and clustering and dividing the remote sensing image data:
s31, determining the quality coefficient of cluster division in the current iteration by using the uncertainty of the cluster division in the feature space in the current iteration and the obtained conditional information entropy of the feature vector set about the remote sensing image data;
s32, according to the obtained quality coefficients, respectively measuring uncertainty of a set S consisting of the remote sensing image data and any two categories in the current iterative clustering division in a symbol space, determining information gain of the set S in the symbol space, combining the two categories in the set S with the minimum information gain into one category, and generating new clustering division;
s33, judging whether the number of the new cluster classification categories is consistent with the number to be classified of the remote sensing images, if so, executing S4, otherwise, returning to S31 to perform the next iteration;
and S4, outputting the final clustering division as a ground feature classification result.
The remote sensing image surface feature classification method measures uncertainty in remote sensing image data by condition information entropy, can effectively reflect uncertainty of information between surface feature information and spectral data in the remote sensing image generation process, and evaluates consistency of remote sensing image surface feature classification information description in a feature space and a symbol space through iterative computation, so that cluster division of the remote sensing image surface feature target is realized, and a surface feature classification result has higher classification accuracy and robustness.
In this embodiment, taking the remote sensing image of a certain coastal zone area obtained by a certain commercial satellite to perform ground object classification as an example, the method for classifying ground objects of remote sensing images provided by the embodiment of the present invention is described, as shown in fig. 2, specifically including the following steps:
s1, acquiring a remote sensing image to be processed, and preprocessing the remote sensing image to obtain a feature vector set of the remote sensing image to be processed, which may specifically include the following steps:
s11, obtaining the remote sensing image to be processed, segmenting an interest region according to the ground feature classification task, and extracting spectral data of a target spectral band combination in the remote sensing image;
in this embodiment, some parameters related to the acquisition of remote sensing image data by a commercial satellite are shown in table 1.
TABLE 1 partial parameters involved in the acquisition of remote sensing image data by commercial satellites
Parameter(s) Index (I)
Obtaining time 4/10/2007
Cloud cover 0％
Image numbering 1010010005911E02
Full color resolution 0.73m
Multispectral resolution 2.92m
Angle of incidence 25°
In the embodiment, remote sensing image data acquired by the commercial satellite is loaded into a computer to be used as a remote sensing image to be processed, a coastal zone area related to the acquired remote sensing image is assumed to belong to an international boundary where airplanes are prohibited from flying, one territory of China in a coverage area of the remote sensing image is selected as a processing area (namely, an interest area), and the geographic positions are 124 degrees 7 '49' E-124 degrees 13 '51' E and 39 degrees 51 '19' N-39 degrees 54 '57' N; and extracting spectral data of 4 spectral band combinations of blue, green, red and near infrared in the remote sensing image.
In this embodiment, it can be determined by looking up literature data that the remote sensing image includes 9 types of coverage of urban houses, roads, industrial sites, rivers, reed beaches, mud beaches, rural houses, agricultural sites, and reservoirs, so that the number of to-be-classified remote sensing images is 9.
S12, extracting the characteristic vector of the remote sensing image to be processed according to the obtained interest area and the extracted spectral data and forming a set
Figure BDA0002423282950000101
Where M is the number of extracted feature vectors, fjRepresenting the jth feature vector, the feature vectors in set F including: spectral characteristics, texture characteristics, average reflection intensity of ground objects in the remote sensing image expressed by a gray mean value, and dispersion degree of the gray value of each pixel of the remote sensing image expressed by a gray median value and the sum of the average values;
in this embodiment, a principal component analysis method may be used to extract spectral features, a gray level co-occurrence matrix method may be used to extract texture features, and then basic statistical analysis may be performed, where the basic statistical analysis includes: and expressing the average reflection intensity of the ground object in the remote sensing image by using the mean value of the gray levels, and reflecting the dispersion degree of the sum of the gray values of all pixels of the remote sensing image and the mean value by using the median value of the gray levels.
S13, performing image enhancement processing based on linear stretching on the remote sensing image data of the interest area, improving the contrast and image definition between the ground object target and the background, and representing the remote sensing image data obtained after the image enhancement processing as
Figure BDA0002423282950000111
Wherein x isiThe image data acquisition method comprises the steps of representing the ith pixel in remote sensing image data, wherein N represents the number of pixels in the image data;
s14, feature mapping: using sets of Gaussian kernel function pairs
Figure BDA0002423282950000112
Performing feature mapping to obtain a mapped feature vector set
Figure BDA0002423282950000113
And to aggregate the feature vectors
Figure BDA0002423282950000114
The feature vectors in (1) obey gaussian distribution; wherein, the Gaussian kernel function is shown as formula (1):
Figure BDA0002423282950000115
wherein κ represents a Gaussian kernel function, xiAnd xjRespectively represent any two pixels in the remote sensing image data X,
Figure BDA0002423282950000116
respectively representing picture elements xi、xjFeature vector collection
Figure BDA0002423282950000117
The value of (a) is selected from,
Figure BDA0002423282950000118
representing a pixel xiAnd xjFeature vector collection
Figure BDA0002423282950000119
Correlation of the above values, γ represents | | | xi(F)-xj(F)||2Wherein, | xi(F)-xj(F)||2Denotes xi(F) And xj(F) In betweenSquare of Euclidean distance, xi(F) Representing a pixel xiValues, x, over a set of characteristic vectors Fi(fj) Representing a pixel xiThe value on the jth eigenvector.
And S2, calculating the uncertainty of the image data in the feature space: determining a set of feature vectors
Figure BDA00024232829500001110
Entropy of conditional information on remote-sensing image data X
Figure BDA00024232829500001111
Describing the uncertainty of the remote sensing image data in the feature space; wherein the content of the first and second substances,
Figure BDA00024232829500001112
the calculation method is as follows:
Figure BDA00024232829500001113
wherein the content of the first and second substances,
Figure BDA00024232829500001114
set of feature vectors representing remote sensing image data X
Figure BDA00024232829500001115
The variance of (c) is calculated by equation (3):
Figure BDA00024232829500001116
wherein, muXFor remote-sensing image data X
Figure BDA00024232829500001117
The method of (a) is not particularly limited,
Figure BDA00024232829500001118
calculated from equation (4):
Figure BDA0002423282950000121
wherein x isa、xbAnd xcRepresenting X in remote-sensing image data XiAny three other picture elements.
S3, performing iteration S31-S33, and clustering and dividing the remote sensing image data:
in this embodiment, before performing the iteration S31-S33, the remote sensing image data X needs to be initialized and clustered, and the specific initialization method is as follows: each picture element is divided into a separate class, i.e.
Figure BDA0002423282950000122
Wherein, C(0)It is indicated that the cluster partitioning is initialized,
Figure BDA0002423282950000123
representing the nth category in the initial cluster partition.
S31, determining a quality coefficient of cluster partitioning in the current iteration by using uncertainty of cluster partitioning in the feature space in the current iteration and the obtained conditional information entropy of the feature vector set about the remote sensing image data, which may specifically include the following steps:
s311, if the current iteration is the mth iteration, calculating a characteristic vector set
Figure BDA0002423282950000124
Clustering division C for remote sensing image data X in current iteration(m)Conditional information entropy of
Figure BDA0002423282950000125
To describe the uncertainty of the cluster partition in the feature space; wherein the content of the first and second substances,
Figure BDA0002423282950000126
the calculation method is as shown in formula (5):
Figure BDA0002423282950000127
wherein the content of the first and second substances,
Figure BDA0002423282950000128
is represented by C(m)Feature vector collection
Figure BDA0002423282950000129
The variance of (c) is calculated by equation (6):
Figure BDA00024232829500001210
wherein x iseIs any pixel in the remote sensing image data X,
Figure BDA00024232829500001211
is C(m)Feature vector collection
Figure BDA00024232829500001212
In the above-mentioned manner, the expectation is that,
Figure BDA00024232829500001213
calculated from equation (7):
Figure BDA00024232829500001214
wherein x isf，xgAnd xhFor X in remote-sensing image data XeAny other 3 pixels;
s312, calculating the characteristic space information gain: according to the obtained characteristic vector set
Figure BDA00024232829500001215
Entropy of conditional information on remote-sensing image data X
Figure BDA00024232829500001216
And feature vector sets
Figure BDA00024232829500001217
Partition C with respect to clustering(m)Conditional information entropy of
Figure BDA00024232829500001218
Determining information gain of cluster partition of remote sensing image data X in characteristic space in current iteration
Figure BDA0002423282950000131
Wherein the content of the first and second substances,
Figure BDA0002423282950000132
the calculation is made from equation (8):
Figure BDA0002423282950000133
s313, calculating a cluster partition quality coefficient: determining the quality coefficient omega of cluster division in the current iteration according to the information gain of the cluster division of the remote sensing image data X in the current iteration in the characteristic space(m)Wherein, ω is(m)The calculation method of (2) is shown in formula (9):
Figure BDA0002423282950000134
wherein, ω is(m)Representing the cluster partition C of the remote-sensing image data X generated by the mth iteration(m)Mass coefficient of (C)(t)And (3) representing the clustering division of the remote sensing image data X generated by the t-th iteration, wherein t is more than or equal to 1 and less than or equal to m.
S32, determining information gain of the set S in the symbol space by measuring uncertainty in the symbol space of the set S composed of the remote sensing image data and any two categories in the current iterative cluster partition, respectively, according to the obtained quality coefficient, and combining the two categories in the set S with the minimum information gain into one category to generate a new cluster partition, which may specifically include the following steps:
s321, representing the cluster division of the remote sensing image data in the current iteration as a cluster vector in a symbol space to obtain a cluster vector set, which may specifically include the following steps:
a1, if it is the m-th iterationAnd representing the clustering division of the remote sensing image data X generated by the mth iteration into a set formed by all categories, and recording the set as the
Figure BDA0002423282950000135
Wherein, C(m)For cluster partitioning in the mth iteration,
Figure BDA0002423282950000136
is represented by C(m)Of the kth class, KmIs represented by C(m)The number of categories in (1);
a2, dividing the cluster of the remote sensing image data X generated by the mth iteration into C in the symbol space(m)Expressed as a cluster vector
Figure BDA0002423282950000137
Thereby, the clustering division of the remote sensing image data is regarded as a representation form in a symbol space, whereinm,kRepresenting a clustering vector lmThe k component of (1)<k<Km；
A3, recording the set formed by a series of clustering vectors generated in the symbol space by the 1 st to m th iterations as a set
Figure BDA0002423282950000138
Wherein ltAnd representing the clustering vectors which are generated by the t iteration and correspond to the clustering division of the remote sensing image data X.
S322, determining an information gain of the set S in the symbol space by respectively measuring the uncertainty of the set S in the symbol space, the set being composed of any two categories in the remote sensing image data and the current iterative clustering partition, according to the obtained quality coefficient and clustering vector set, and may specifically include the following steps:
b1, if the current iteration is the m-th iteration, determining a cluster vector set in the m-th iteration
Figure BDA0002423282950000141
The conditional information entropy of each cluster vector about the remote sensing image data X is described to describe the remote sensing image data X in each clusterUncertainty on the class vector, where for 1 ≦ t ≦ m,
Figure BDA0002423282950000142
middle t clustering vector ltConditional information entropy H (l) of remote sensing image data Xt| X) is calculated by equation (10):
Figure BDA0002423282950000143
wherein, KtIs represented by C(t)Number of classes in (i.e. clustering vector l)tThe number of components of (a); lt,kAs a cluster vector ltThe k component of (a), P (l)t,k| X) represents lt,kThe conditional probability of the remote sensing image data X is calculated by equation (11):
Figure BDA0002423282950000144
wherein x isi(lt) Representing a pixel xiIn the clustering vector ltThe value of (a), i.e. pixel x in the t-th iterationiThe category into which it is divided;
b2, determining a cluster vector set in the mth iteration
Figure BDA0002423282950000145
Entropy of conditional information on remote-sensing image data X
Figure BDA0002423282950000146
To describe the uncertainty in the symbol space of the remotely sensed image data X, wherein,
Figure BDA0002423282950000147
calculated from equation (12):
Figure BDA0002423282950000148
wherein, ω is(t)Representing the cluster partition C produced by the t-th iteration(t)The mass coefficient of (a);
b3, partition in clusters C(m)Optionally two of
Figure BDA0002423282950000149
And
Figure BDA00024232829500001410
1≤p≠q≤Kmcomputing a set of clustering vectors
Figure BDA00024232829500001411
About a set for each cluster vector
Figure BDA00024232829500001412
To describe the uncertainty of the set S of these two classes on each cluster vector, wherein,
Figure BDA00024232829500001413
t th clustering vector ltAbout
Figure BDA00024232829500001414
Conditional information entropy of
Figure BDA00024232829500001415
Calculated from equation (13):
Figure BDA00024232829500001416
wherein the content of the first and second substances,
Figure BDA00024232829500001417
representing a clustering vector ltWith respect to the set
Figure BDA00024232829500001418
The conditional probability of (a) of (b),
Figure BDA0002423282950000151
calculated from equation (14):
Figure BDA0002423282950000152
b4, determining a cluster vector set
Figure BDA0002423282950000153
About collections
Figure BDA0002423282950000154
Conditional information entropy of
Figure BDA0002423282950000155
To describe a collection
Figure BDA0002423282950000156
Uncertainty in the symbol space, wherein,
Figure BDA0002423282950000157
calculated from equation (15):
Figure BDA0002423282950000158
b5, calculating symbol space information gain: determining a set
Figure BDA0002423282950000159
Information gain in symbol space
Figure BDA00024232829500001510
Calculated from equation (16):
Figure BDA00024232829500001511
s323, pixel category fusion: merging two categories in the set S with minimum information gain into one category to generate a new cluster partition C(m+1)。
S33, judging whether the number of the new cluster classification categories is consistent with the number to be classified of the remote sensing images, if so, executing S4, otherwise, returning to S31 to perform the next iteration;
and S4, outputting the final clustering division as a ground feature classification result.
In this embodiment, the cluster division of the remote sensing image data is output as a surface feature classification result, that is, each pixel class in the cluster division corresponds to an actual surface feature target, and this surface feature classification result is used to help an analyst of the remote sensing image to perform various operations such as surface feature target identification, surface change identification, and Geographic Information System (GIS) database update.
Finally, in order to verify the effectiveness and the advancement of the remote sensing image ground feature classification method provided by the invention, a k-means clustering (k-means) algorithm and an iterative self-organizing data analysis algorithm (ISODATA) are selected to compare the remote sensing image ground feature classification effects. The results of the classification of the surface features of the 3 methods were evaluated by the remote sensing image data in the examples using the average classification accuracy and the classification variance of 20 experiments as evaluation indexes, and the evaluation results are shown in table 2:
TABLE 2 evaluation results
Figure BDA0002423282950000161
The results in table 2 show that the remote sensing image surface feature classification method provided by the invention can obtain higher classification precision and effectively improve the stability of the classification result when the remote sensing image surface feature classification is carried out.
Example two
The present invention further provides a specific embodiment of a remote sensing image land feature classification system, which corresponds to the specific embodiment of the remote sensing image land feature classification method, and the remote sensing image land feature classification system can achieve the object of the present invention by executing the flow steps in the specific embodiment of the method, so the explanation in the specific embodiment of the remote sensing image land feature classification method is also applicable to the specific embodiment of the remote sensing image land feature classification system provided by the present invention, and will not be repeated in the following specific embodiment of the present invention.
As shown in fig. 3, an embodiment of the present invention further provides a remote sensing image ground feature classification system, including:
the processing module 11 is configured to acquire a remote sensing image to be processed, process the remote sensing image to be processed, and obtain a feature vector set of the remote sensing image to be processed;
the determining module 12 is used for calculating the conditional information entropy of the feature vector set about the remote sensing image data to describe the uncertainty of the remote sensing image data in the feature space;
the partitioning module 13 is used for iteratively executing S31-S33, and clustering and partitioning the remote sensing image data:
s31, determining the quality coefficient of cluster division in the current iteration by using the uncertainty of the cluster division in the feature space in the current iteration and the obtained conditional information entropy of the feature vector set about the remote sensing image data;
s32, according to the obtained quality coefficients, respectively measuring uncertainty of a set S consisting of the remote sensing image data and any two categories in the current iterative clustering division in a symbol space, determining information gain of the set S in the symbol space, combining the two categories in the set S with the minimum information gain into one category, and generating new clustering division;
s33, judging whether the number of the new cluster classification categories is consistent with the number to be classified of the remote sensing images, if so, executing the output module 14, otherwise, returning to S31 for next iteration;
and the output module 14 is used for outputting the final clustering division as a ground feature classification result.
The remote sensing image ground object classification system measures uncertainty in remote sensing image data by condition information entropy, can effectively reflect uncertainty of information between ground object information and spectral data in the remote sensing image generation process, and evaluates consistency of remote sensing image ground object classification information description in feature space and symbol space through iterative computation, so that cluster division of ground object targets of remote sensing images is realized, and ground object classification results have higher classification accuracy and robustness.
While the foregoing is directed to the preferred embodiment of the present invention, it will be understood by those skilled in the art that various changes and modifications may be made without departing from the spirit and scope of the invention as defined in the appended claims.

Claims (10)

1. A remote sensing image surface feature classification method is characterized by comprising the following steps:
s1, acquiring the remote sensing image to be processed, and preprocessing the remote sensing image to be processed to obtain a characteristic vector set of the remote sensing image to be processed;
s2, calculating the condition information entropy of the feature vector set about the remote sensing image data to describe the uncertainty of the remote sensing image data in the feature space;
s3, performing iteration S31-S33, and clustering and dividing the remote sensing image data:
s31, determining the quality coefficient of cluster division in the current iteration by using the uncertainty of the cluster division in the feature space in the current iteration and the obtained conditional information entropy of the feature vector set about the remote sensing image data;
s32, according to the obtained quality coefficients, respectively measuring uncertainty of a set S consisting of the remote sensing image data and any two categories in the current iterative clustering division in a symbol space, determining information gain of the set S in the symbol space, combining the two categories in the set S with the minimum information gain into one category, and generating new clustering division;
s33, judging whether the number of the new cluster classification categories is consistent with the number to be classified of the remote sensing images, if so, executing S4, otherwise, returning to S31 to perform the next iteration;
and S4, outputting the final clustering division as a ground feature classification result.
2. The remote sensing image land feature classification method according to claim 1, wherein the obtaining and preprocessing the remote sensing image to be processed to obtain the feature vector set of the remote sensing image to be processed comprises:
obtaining a remote sensing image to be processed, segmenting an interest area according to a ground object classification task, and extracting spectral data of a target spectral band combination in the remote sensing image;
extracting the characteristic vector of the remote sensing image to be processed according to the obtained interest area and the extracted spectral data and forming a set
Figure FDA0002423282940000011
Where M is the number of extracted feature vectors, fjRepresenting the jth feature vector, the feature vectors in set F including: spectral characteristics, texture characteristics, average reflection intensity of ground objects in the remote sensing image expressed by a gray mean value, and dispersion degree of the gray value of each pixel of the remote sensing image expressed by a gray median value and the sum of the average values;
carrying out image enhancement processing based on linear stretching on the remote sensing image data of the interest area to obtain the remote sensing image data
Figure FDA0002423282940000021
Wherein x isiThe image data acquisition method comprises the steps of representing the ith pixel in remote sensing image data, wherein N represents the number of pixels in the image data;
using sets of Gaussian kernel function pairs
Figure FDA0002423282940000022
Performing feature mapping to obtain a mapped feature vector set
Figure FDA0002423282940000023
And to aggregate the feature vectors
Figure FDA0002423282940000024
The feature vectors in (1) obey gaussian distribution; wherein the gaussian kernel function is represented as:
Figure FDA0002423282940000025
wherein κ represents a Gaussian kernel function, xiAnd xjRespectively represent any two pixels in the remote sensing image data X,
Figure FDA0002423282940000026
respectively representing picture elements xi、xjFeature vector collection
Figure FDA0002423282940000027
The value of (a) is selected from,
Figure FDA0002423282940000028
representing a pixel xiAnd xjFeature vector collection
Figure FDA0002423282940000029
Correlation of the above values, γ represents | | | xi(F)-xj(F)||2Standard deviation of (1), xi(F) Representing a pixel xiValues, x, over a set of characteristic vectors Fi(fj) Representing a pixel xiThe value on the jth eigenvector.
3. The remote sensing image land feature classification method according to claim 2, wherein the entropy of the condition information of the feature vector set on the remote sensing image data is expressed as:
Figure FDA00024232829400000210
wherein the content of the first and second substances,
Figure FDA00024232829400000211
representing a set of feature vectors
Figure FDA00024232829400000212
Entropy of condition information about the remote sensing image data X;
Figure FDA00024232829400000213
set of feature vectors representing remote sensing image data X
Figure FDA00024232829400000214
The variance of (a) above (b),
Figure FDA00024232829400000215
the calculation formula of (2) is as follows:
Figure FDA00024232829400000216
wherein, muXRepresenting remote-sensed image data X in
Figure FDA00024232829400000217
The method of (a) is not particularly limited,
Figure FDA00024232829400000218
the calculation formula of (2) is as follows:
Figure FDA00024232829400000219
wherein x isa、xbAnd xcRepresenting X in remote-sensing image data XiAny three other picture elements.
4. The remote sensing image land feature classification method according to claim 1, before iteratively performing S31-S33, the method further comprising:
carrying out initialization clustering division on remote sensing image data: each picture element is divided into a separate class, i.e.
Figure FDA0002423282940000031
Wherein, C(0)It is indicated that the cluster partitioning is initialized,
Figure FDA0002423282940000032
indicates the beginningBegin clustering the nth category in the partition.
5. The method for classifying terrestrial objects according to claim 3, wherein the calculating the quality coefficient of cluster partitioning in the current iteration by using the uncertainty of cluster partitioning in the feature space in the current iteration and the conditional information entropy of the obtained feature vector set on the remote sensing image data comprises:
if the current iteration is the m-th iteration, calculating a characteristic vector set
Figure FDA0002423282940000033
Clustering division C for remote sensing image data X in current iteration(m)Conditional information entropy of
Figure FDA0002423282940000034
To describe the uncertainty of the cluster partition in the feature space; wherein the content of the first and second substances,
Figure FDA0002423282940000035
expressed as:
Figure FDA0002423282940000036
wherein the content of the first and second substances,
Figure FDA0002423282940000037
is represented by C(m)Feature vector collection
Figure FDA0002423282940000038
The variance of (a) above (b),
Figure FDA0002423282940000039
the calculation formula of (2) is as follows:
Figure FDA00024232829400000310
wherein the content of the first and second substances,xeis any pixel in the remote sensing image data X,
Figure FDA00024232829400000311
is C(m)Feature vector collection
Figure FDA00024232829400000312
In the above-mentioned manner, the expectation is that,
Figure FDA00024232829400000313
the calculation formula of (2) is as follows:
Figure FDA00024232829400000314
wherein x isf，xgAnd xhFor X in remote-sensing image data XeAny other 3 pixels;
according to the obtained characteristic vector set
Figure FDA00024232829400000315
Entropy of conditional information on remote-sensing image data X
Figure FDA00024232829400000316
And feature vector sets
Figure FDA00024232829400000317
Partition C with respect to clustering(m)Conditional information entropy of
Figure FDA00024232829400000318
Determining information gain of cluster partition of remote sensing image data X in characteristic space in current iteration
Figure FDA00024232829400000319
Wherein the content of the first and second substances,
Figure FDA00024232829400000320
expressed as:
Figure FDA00024232829400000321
and determining the quality coefficient of cluster division in the current iteration according to the information gain of the cluster division of the remote sensing image data X in the current iteration in the characteristic space.
6. The method of claim 5, wherein the mth iteration generates cluster partition C for the remote sensing image data X(m)Mass coefficient of (omega)(m)Expressed as:
Figure FDA0002423282940000041
wherein, C(t)And (3) representing the clustering division of the remote sensing image data X generated by the t-th iteration, wherein t is more than or equal to 1 and less than or equal to m.
7. The method for classifying terrestrial objects according to claim 6, wherein the determining an information gain of a set S in a symbol space by measuring uncertainty of the set S consisting of the remote-sensing image data and any two categories in the current iterative clustering partition in the symbol space respectively according to the obtained quality coefficients, and combining the two categories in the set S with the minimum information gain into one category to generate a new clustering partition comprises:
clustering division of the remote sensing image data in the current iteration is expressed as a clustering vector in a symbol space, and a clustering vector set is obtained;
according to the obtained quality coefficient and the cluster vector set, respectively measuring the uncertainty of a set S consisting of any two categories in the remote sensing image data and the current iterative clustering division in a symbol space, and determining the information gain of the set S in the symbol space;
and combining two categories in the set S with the minimum information gain into one category to generate a new cluster partition.
8. The method for classifying terrestrial objects according to claim 7, wherein the step of dividing clusters of remote-sensing image data in a current iteration into cluster vectors in a symbol space to obtain a cluster vector set comprises:
if the current iteration is the mth iteration, the cluster division of the remote sensing image data X generated by the mth iteration is expressed as a set formed by all the classified categories and recorded as
Figure FDA0002423282940000042
Wherein, C(m)For cluster partitioning in the mth iteration,
Figure FDA0002423282940000043
is represented by C(m)Of the kth class, KmIs represented by C(m)The number of categories in (1);
dividing the cluster of the remote sensing image data X generated by the mth iteration into C in the symbol space(m)Expressed as a cluster vector
Figure FDA0002423282940000044
Wherein lm,kRepresenting a clustering vector lmThe k component of (1)<k<Km；
Recording a set formed by a series of clustering vectors generated in the symbol space by the 1 st to m iterations as a set
Figure FDA0002423282940000045
Wherein ltAnd representing the clustering vectors which are generated by the t iteration and correspond to the clustering division of the remote sensing image data X.
9. The method for classifying terrestrial objects according to claim 8, wherein the determining, according to the obtained quality coefficient and cluster vector set, an information gain of a set S in a symbol space by separately measuring uncertainty of the set S in the symbol space, the set S being composed of any two categories in the remote sensing image data and the current iterative cluster partition, comprises:
if the current iteration is the m-th iteration, determining a cluster vector set in the m-th iteration
Figure FDA0002423282940000051
Wherein each cluster vector is associated with an entropy of conditional information of the remote sensing image data X to describe an uncertainty of the remote sensing image data X in each cluster vector, wherein for t ≦ 1 ≦ m,
Figure FDA0002423282940000052
middle t clustering vector ltConditional information entropy H (l) of remote sensing image data Xt| X) is expressed as:
Figure FDA0002423282940000053
wherein, KtIs represented by C(t)Number of classes in, lt,kAs a cluster vector ltThe k component of (a), P (l)t,k| X) represents lt,kConditional probability, P (l), for remote-sensed image data Xt,k| X) is calculated as:
Figure FDA0002423282940000054
wherein x isi(lt) Representing a pixel xiIn the clustering vector ltThe value of (a), i.e. pixel x in the t-th iterationiThe category into which it is divided;
determining a set of clustering vectors in an mth iteration
Figure FDA0002423282940000055
Entropy of conditional information on remote-sensing image data X
Figure FDA0002423282940000056
To describe the uncertainty in the symbol space of the remotely sensed image data X, wherein,
Figure FDA0002423282940000057
expressed as:
Figure FDA0002423282940000058
wherein, ω is(t)Representing the cluster partition C produced by the t-th iteration(t)The mass coefficient of (a);
partition C in clustering(m)Optionally two of
Figure FDA0002423282940000059
And
Figure FDA00024232829400000510
1≤p≠q≤Kmcomputing a set of clustering vectors
Figure FDA00024232829400000511
About a set for each cluster vector
Figure FDA00024232829400000512
To describe the uncertainty of the set S of these two classes on each cluster vector, wherein,
Figure FDA00024232829400000513
t th clustering vector ltAbout
Figure FDA00024232829400000514
Conditional information entropy of
Figure FDA00024232829400000515
Expressed as:
Figure FDA00024232829400000516
wherein the content of the first and second substances,
Figure FDA0002423282940000061
representing a clustering vector ltWith respect to the set
Figure FDA0002423282940000062
The conditional probability of (a) of (b),
Figure FDA0002423282940000063
expressed as:
Figure FDA0002423282940000064
determining a set of clustering vectors
Figure FDA0002423282940000065
About collections
Figure FDA0002423282940000066
Conditional information entropy of
Figure FDA0002423282940000067
To describe a collection
Figure FDA0002423282940000068
Uncertainty in the symbol space, wherein,
Figure FDA0002423282940000069
expressed as:
Figure FDA00024232829400000610
determining a set
Figure FDA00024232829400000611
Information gain in symbol space
Figure FDA00024232829400000612
Figure FDA00024232829400000613
Expressed as:
Figure FDA00024232829400000614
10. a remote sensing image surface feature classification system is characterized by comprising:
the processing module is used for acquiring the remote sensing image to be processed and processing the remote sensing image to be processed to obtain a characteristic vector set of the remote sensing image to be processed;
the determining module is used for determining the conditional information entropy of the feature vector set about the remote sensing image data so as to describe the uncertainty of the remote sensing image data in the feature space;
the partitioning module is used for iteratively executing S31-S33 and clustering and partitioning the remote sensing image data:
s31, determining the quality coefficient of cluster division in the current iteration by using the uncertainty of the cluster division in the feature space in the current iteration and the obtained conditional information entropy of the feature vector set about the remote sensing image data;
s32, according to the obtained quality coefficients, respectively measuring uncertainty of a set S consisting of the remote sensing image data and any two categories in the current iterative clustering division in a symbol space, determining information gain of the set S in the symbol space, combining the two categories in the set S with the minimum information gain into one category, and generating new clustering division;
s33, judging whether the number of the new cluster classification categories is consistent with the number to be classified of the remote sensing images, if so, executing an output module, otherwise, returning to S31 for next iteration;
and the output module is used for outputting the final clustering division as a ground feature classification result.
CN202010212440.9A 2020-03-24 2020-03-24 Remote sensing image ground object classification method and system Pending CN111460943A (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202010212440.9A CN111460943A (en) 2020-03-24 2020-03-24 Remote sensing image ground object classification method and system

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202010212440.9A CN111460943A (en) 2020-03-24 2020-03-24 Remote sensing image ground object classification method and system

Publications (1)

Publication Number Publication Date
CN111460943A true CN111460943A (en) 2020-07-28

Family

ID=71679802

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202010212440.9A Pending CN111460943A (en) 2020-03-24 2020-03-24 Remote sensing image ground object classification method and system

Country Status (1)

Country Link
CN (1) CN111460943A (en)

Cited By (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN113537329A (en) * 2021-07-30 2021-10-22 山西大学 Method for rapidly estimating probability distribution of various ground objects position by position
CN113554675A (en) * 2021-07-19 2021-10-26 贵州师范大学 Edible fungus yield estimation method based on unmanned aerial vehicle visible light remote sensing

Citations (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN104376335A (en) * 2014-11-13 2015-02-25 河南理工大学 Semi-supervised hyperspectral remote sensing image classification method based on information entropies
CN109034238A (en) * 2018-07-23 2018-12-18 山西大学 A kind of clustering method based on comentropy
CN109657616A (en) * 2018-12-19 2019-04-19 四川立维空间信息技术有限公司 A kind of remote sensing image land cover pattern automatic classification method

Patent Citations (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN104376335A (en) * 2014-11-13 2015-02-25 河南理工大学 Semi-supervised hyperspectral remote sensing image classification method based on information entropies
CN109034238A (en) * 2018-07-23 2018-12-18 山西大学 A kind of clustering method based on comentropy
CN109657616A (en) * 2018-12-19 2019-04-19 四川立维空间信息技术有限公司 A kind of remote sensing image land cover pattern automatic classification method

Non-Patent Citations (4)

* Cited by examiner, † Cited by third party
Title
BEGÜM DEMIR等: "Detection of Land-Cover Transitions in Multitemporal Remote Sensing Images With Active-Learning-Based Compound Classification", 《IEEE TRANSACTIONS ON GEOSCIENCE AND REMOTE SENSING》 *
LIANG BAI等: "An Information-Theoretical Framework for Cluster Ensemble", 《IEEE TRANSACTIONS ON KNOWLEDGE AND DATA ENGINEERING》 *
张盈等: "顾及空间相关性的遥感影像信息量的度量方法", 《测绘学报》 *
黄冬梅等: "利用信息熵的高光谱遥感影像降维方法", 《计算机工程与应用》 *

Cited By (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN113554675A (en) * 2021-07-19 2021-10-26 贵州师范大学 Edible fungus yield estimation method based on unmanned aerial vehicle visible light remote sensing
CN113537329A (en) * 2021-07-30 2021-10-22 山西大学 Method for rapidly estimating probability distribution of various ground objects position by position

Similar Documents

Publication Publication Date Title
Zhang et al. A feature difference convolutional neural network-based change detection method
Shendryk et al. Deep learning for multi-modal classification of cloud, shadow and land cover scenes in PlanetScope and Sentinel-2 imagery
CN103971115B (en) Automatic extraction method for newly-increased construction land image spots based on NDVI and PanTex index
CN109871902B (en) SAR small sample identification method based on super-resolution countermeasure generation cascade network
CN110097101B (en) Remote sensing image fusion and coastal zone classification method based on improved reliability factor
CN109063754B (en) Remote sensing image multi-feature joint classification method based on OpenStreetMap
Mason et al. The use of digital map data in the segmentation and classification of remotely-sensed images
CN110349160A (en) One kind is based on super-pixel and fuzzy C-means clustering SAR image segmentation method
CN110751087B (en) EOF-based unmanned aerial vehicle signal identification system and method
Guan et al. Random forests-based feature selection for land-use classification using lidar data and orthoimagery
CN111460943A (en) Remote sensing image ground object classification method and system
Khoshboresh-Masouleh et al. A deep learning method for near-real-time cloud and cloud shadow segmentation from gaofen-1 images
Asiyabi et al. Earth observation semantic data mining: Latent dirichlet allocation-based approach
Zhang et al. A novel multitemporal cloud and cloud shadow detection method using the integrated cloud Z-scores model
Seydi et al. A quadratic morphological deep neural network fusing radar and optical data for the mapping of burned areas
Liu Advanced techniques for automatic change detection in multitemporal hyperspectral images
Liu et al. A dual-channel fully convolutional network for land cover classification using multifeature information
Sima et al. Sparse representation classification based on flexible patches sampling of superpixels for hyperspectral images
Hong et al. Random forest fusion classification of remote sensing PolSAR and optical image based on LASSO and IM factor
Chen et al. Equivalent complex valued deep semantic segmentation network for sar images
Liu et al. Semi-automatic extraction and mapping of farmlands based on high-resolution remote sensing images
Liu et al. A foreground/background separation framework for interpreting polarimetric SAR images
Nuradili et al. UAV Remote Sensing Image Semantic Segmentation Strategy Based on Thermal Infrared and Multispectral Image Features
Yang et al. AAE-Dpeak-SC: A novel unsupervised clustering method for space target ISAR images based on adversarial autoencoder and density peak-spectral clustering
Gressin et al. A unified framework for land-cover database update and enrichment using satellite imagery

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
RJ01 Rejection of invention patent application after publication

Application publication date: 20200728

RJ01 Rejection of invention patent application after publication