CN107203791B - Polarimetric SAR image classification method based on heteropolarity ratio and scattering power entropy - Google Patents

Polarimetric SAR image classification method based on heteropolarity ratio and scattering power entropy Download PDF

Info

Publication number
CN107203791B
CN107203791B CN201710522622.4A CN201710522622A CN107203791B CN 107203791 B CN107203791 B CN 107203791B CN 201710522622 A CN201710522622 A CN 201710522622A CN 107203791 B CN107203791 B CN 107203791B
Authority
CN
China
Prior art keywords
scattering
sar image
polarized
ratio
scattering power
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
CN201710522622.4A
Other languages
Chinese (zh)
Other versions
CN107203791A (en
Inventor
王爽
焦李成
周小凤
滑文强
段丽英
赵阳
侯彪
马文萍
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.)
Xian University of Electronic Science and Technology
Original Assignee
Xian University of Electronic Science and Technology
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 Xian University of Electronic Science and Technology filed Critical Xian University of Electronic Science and Technology
Priority to CN201710522622.4A priority Critical patent/CN107203791B/en
Publication of CN107203791A publication Critical patent/CN107203791A/en
Application granted granted Critical
Publication of CN107203791B publication Critical patent/CN107203791B/en
Active legal-status Critical Current
Anticipated expiration legal-status Critical

Links

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F18/00—Pattern recognition
    • G06F18/20—Analysing
    • G06F18/24—Classification techniques
    • G06F18/243—Classification techniques relating to the number of classes
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
    • G06V10/00—Arrangements for image or video recognition or understanding
    • G06V10/20—Image preprocessing
    • G06V10/30—Noise filtering

Abstract

The invention provides a polarimetric SAR image classification method based on a heteropolarity ratio and a scattering power entropy, which is used for solving the technical problem of low classification precision in the existing unsupervised polarimetric SAR image classification method. The method comprises the following implementation steps: removing speckle noise in the polarized SAR image to be classified; performing Freeman decomposition on the polarized SAR image to obtain three scattering powers of the image; calculating the scattering power entropy of the image according to the three kinds of scattering power; initially dividing the polarized SAR image into 7 classes by using three scattering powers and scattering power entropies; calculating the ratio of the co-polarization component to the cross-polarization component of each pixel point in the polarized SAR image: the heteropolarity ratio; carrying out subdivision on each class in the 7 classes of polarized SAR data in the same proportion by using the ratio; merging the classification results based on a specific inter-class merging criterion; and performing complex Wishart iteration on the combined result and coloring to obtain a final color classification diagram.

Description

Polarimetric SAR image classification method based on heteropolarity ratio and scattering power entropy
Technical Field
The invention belongs to the technical field of image data processing, relates to a polarized SAR image classification method, and particularly relates to a polarized SAR image classification method based on a heteropolarity ratio and a scattering power entropy, which can be used for ground feature classification of a polarized SAR image.
Background
The polarimetric synthetic aperture radar is a multi-parameter and multi-channel radar imaging system, and has been widely researched and applied in the fields of agriculture, military, ocean and the like due to the strong target information acquisition capability, and has become one of the important directions of the development of the synthetic aperture radar, and the polarimetric SAR image classification is one of the important applications of the polarimetric synthetic aperture radar and receives more and more attention. The polarized SAR image classification is to classify each pixel in the image by utilizing polarized measurement data provided by an airborne or satellite-borne polarized SAR system, thereby realizing the classification of the polarized SAR image. The existing polarized SAR image classification method mainly comprises an unsupervised classification method and a supervised classification method, but the supervised classification method needs a large number of training samples, and the classification precision is easily influenced by training data. In the process of performing terrain classification on the images, in many cases, the real data of the ground surface is unknown, so that the artificial selection of training samples is difficult for polarizing the SAR images. Therefore, the feature classification at present focuses on unsupervised classification, and the feature of unsupervised classification is that training samples are not needed, the real data distribution of features in images is not needed to be known, and data information contained in the images is fully utilized to classify the images. The method for comparing the unsupervised classification of the classical polarized SAR image comprises the following steps: on the basis of these classical classification algorithms, classification strategies of first classification and then combination are proposed, which include:
in 2004, j.s.lee et al proposed an Unsupervised classification algorithm for polarized SAR images based on Freeman decomposition, see j.s.lee, m.r.gruns, e.pottier and l.ferro-family, "Unsupervised tertiary in classification modeling prediction polar scattering characteristics," in IEEE Transactions on Geoscience and Remote Sensing, vol.42, No.4, pp.722-731, april2004. this method decomposes images into three categories, surface scattering, dihedral scattering and volume scattering, based on the Freeman scattering model, then subdivides each category into 30 categories or more according to the magnitude of the scattering power of each pixel in each category of data, and then performs category merging and wiskart iteration on the data. The method combines a Freeman scattering model and complex Wishart iteration, has the characteristic of keeping the purity of a main scattering mechanism of the multi-polarization SAR image, but does not consider the existence of a mixed scattering mechanism in the polarization SAR image, so the classification precision is still to be improved.
In 2007, Cao et al proposed a classification method of adaptive class polarized SAR images based on SPAN/H/α/a and complex Wishart algorithm, see f.cao, w.hong, y.wu, and e.pottier as "An adaptive number of clusters using the SPAN/H/α/a space and the complex Wishart clustering for a full polar SAR analysis, which combines the backscatter power SPAN data of data and H/α/a information of data, classify polarized images into 48 classes, then use a Wishart test statistical method to realize a clustering type of classification of data, and evaluate the likelihood of data being suitable for a hierarchical classification of data. The method can perform self-adaptive classification on the polarized SAR data, the robustness of the method is guaranteed, but the method still does not consider the existence of a mixed scattering mechanism in the polarized SAR image, and the method uses the aggregation hierarchical clustering and the likelihood evaluation of each clustering on the data, so that the calculation complexity of the whole method is higher.
In 2013, Wang et al proposed Classification methods Based on Scattering Power Entropy and co-polarization Ratio, see S.Wang, K.Liu, J.Pei, M.Gong and Y.Liu, "Unsupervised Classification of full polar spectra Based on Scattering Power Entry and polarized Ratio," in IEEEGeoscity and Remote Sensing Letters, vol.10, No.3, pp.622-626, May 2013. the method decomposes images into seven major classes, surface Scattering class, dihedral Scattering class, surface-bulk Scattering class, dihedral-bulk Scattering class, and mixed class Based on the Scattering Power Entropy, then classifies each major class into three minor classes using the co-polarization Ratio characteristics of the data, and finally, iteratively, classifying the results by a Wishart Classification method. The method considers the existence of a mixed scattering mechanism in the pixel, but the used co-polarization ratio only relates to the co-polarization ratio component in the polarized SAR image, and the cross-polarization component of the polarized SAR image is ignored, so that the problems of partial target misclassification or non-fine classification are caused.
Disclosure of Invention
The invention aims to overcome the defects of the prior art and provides a polarized SAR image classification method based on the heteropolarity ratio and the scattering power entropy so as to realize more accurate ground feature classification of the polarized SAR image.
The technical idea of the invention is as follows: firstly, on the basis of Freeman decomposition, initially classifying polarized SAR images by using scattering power entropy; then, each type of the image initial division is further divided according to the ratio of the co-polarization component and the cross-polarization component of each pixel point in the polarization SAR image; combining the classification results by adopting an inter-class combination criterion based on Wishart distance, and performing Wishart iteration on the combined result; and finally coloring the image to obtain a colored classification result graph, wherein the specific implementation steps are as follows:
(1) removing speckle noise in the polarized SAR image to be classified to obtain a filtered polarized SAR image;
(2) acquiring three scattering powers of each pixel point in the polarized SAR image: performing Freeman decomposition on the filtered polarized SAR image to obtain the surface scattering power P of each pixel pointsDihedral angle scattering power PdAnd bulk scattered power Pv；
(3) Respectively calculating the scattering power entropy H of each pixel point by using the three scattering powers of each pixel pointpWherein 0 is not more than Hp≤1；
(4) Carrying out initial classification on the polarized SAR image: according to three kinds of scattering power and scattering power entropy H of each pixel point in the polarized SAR imagepPerforming initial classification on the polarized SAR imageAnd 7 types of polarized SAR images are obtained:
(4a) scattering power entropy H from polarized SAR imagespSelecting two classification thresholds x1And x2；
(4b) According to the scattering power entropy HpWith two classification thresholds x1And x2The initial classification of the polarized SAR image is specifically as follows:
when H is more than or equal to 0p≤x1Will satisfy Ps＝max(Ps,Pd,Pv) The pixel points are classified into surface scattering classes to satisfy Pd＝max(Ps,Pd,Pv) The pixel points are divided into dihedral angle scattering classes satisfying Pv＝max(Ps,Pd,Pv) The pixel points are divided into volume scattering classes;
when x is1＜Hp≤x2Will satisfy Ps＝min(Ps,Pd,Pv) The pixel points are divided into dihedral angle-volume scattering mixed classes to satisfy Pd＝min(Ps,Pd,Pv) The pixel points are divided into surface-volume scattering mixed classes to satisfy Pv＝min(Ps,Pd,Pv) The pixel points are divided into surface-dihedral angle scattering mixed classes;
when x is2＜HpIf the number of the pixel points is less than or equal to 1, the pixel points are divided into scattering mixed types;
(5) defining a heteropolarity ratio formula, calculating the ratio of a co-polarized component to a cross-polarized component of each pixel point in the polarized SAR image by using the heteropolarity ratio formula to obtain the heteropolarity ratio of each pixel point, and then counting the heteropolarity ratio distribution of the polarized SAR image;
(6) reclassifying the polarized SAR image: according to the heteropolarity ratio of each pixel point in the polarized SAR image and the heteropolarity ratio distribution of the polarized SAR image, carrying out subdivision on each class in the 7 classes of polarized SAR images in the step (4) in the same proportion to obtain 7n classes of polarized SAR images, wherein n represents the class number of subdivision of each class of data, and n is more than or equal to 2;
(7) performing category merging on the obtained 7n types of polarized SAR images to obtain m types of polarized SAR images, wherein the value of m is determined by the number of surface feature types contained in the polarized SAR images;
(8) updating the category of each pixel point in the m-type polarized SAR images to obtain a classification result;
(9) and (4) coloring the classification result obtained in the step (8) according to a three-primary-color principle by using three color components of red R, green G and blue B as three primary colors to obtain a final color classification result graph.
Compared with the prior art, the invention has the following advantages:
the invention initially classifies the polarized SAR images by using the scattering power entropy on the basis of Freeman decomposition, and fully utilizes the polarization information contained in the polarized SAR images on the basis of the initial classification: the co-polarized component and the cross-polarized component divide the data more finely according to the ratio of the two polarized components. The scattering type of each pixel can be effectively determined by the scattering power entropy information, the difference between different ground objects can be well represented by the ratio of the co-polarization component to the cross-polarization component, and the different ground object types can be distinguished conveniently, so that the precision of polarized SAR image classification can be effectively improved by the combination of the ratio of the two polarization components and the scattering power entropy.
Drawings
FIG. 1 is a block diagram of an implementation flow of the present invention;
FIG. 2 is a polarized SAR image San Francisco image used in the simulation of the present invention;
FIG. 3 is a heteropolarity ratio distribution diagram of principal features in a polarized SAR image used in the simulation of the present invention;
FIG. 4 is a graph showing the results of classification simulation of San Francisco images by the present invention and the existing classification method based on scattering power entropy and co-polarization ratio.
Detailed Description
The invention is described in further detail below with reference to the figures and the specific examples.
Referring to fig. 1, a method for classifying a polarized SAR image based on a heteropolarity ratio and a scattering power entropy includes the following steps:
step 1) removing speckle noise in the San Francisco image by adopting a refined Lee filtering method to obtain a filtered San Francisco image;
step 2) carrying out Freeman decomposition on the filtered San Francisco image to obtain the surface scattering power P of each pixel pointsDihedral angle scattering power PdAnd bulk scattered power PvThe method comprises the following implementation steps:
(2a) inputting a covariance matrix C of each pixel point in the San Francisco image:
wherein H denotes horizontal polarization, V denotes vertical polarization, SHHRepresenting echo images received horizontally, SVVRepresenting vertically transmitted and vertically received echo images, SHVRepresenting horizontally transmitted and vertically received echo images, SHHAnd SVVReferred to as co-polarized component, SHVReferred to as cross-polarization components, representing conjugation, |, representing modulus values,<·>represents a per-view average;
(2b) the covariance matrix C is decomposed into the following form:
wherein f issIs the coefficient of decomposition, f, of the surface scattering componentdIs the decomposition coefficient of the dihedral scattering component, fvBeta represents the ratio of the horizontal emission horizontal reception backscatter reflection coefficient to the vertical emission vertical reception backscatter reflection coefficient, and alpha is defined as alpha-RghRvh/RgvRvv，Rgh，RvhRespectively representing the horizontal and vertical reflection coefficients of the earth's surface, Rgv，RvvRespectively representing the horizontal and vertical reflection coefficients of a vertical wall;
(2c) according to the equality of corresponding elements of the formula 1) and the formula 2), five unknowns f are obtaineds，fd，fvα andfour equations for β:
(2d) solving equation set 3) to obtain fs，fd，fvValues of α and β:
calculating the covariance matrix of each pixelIf the positive and negative values are judged, ifLet α be-1, otherwise let β be 1, solve equation set 3) based on the values of α and β, and then obtain f)s，fdAnd fvWherein Re (. cndot.) represents the real part;
(2e) according to fs，fd，fvAnd the values of alpha and beta, and the following formula is utilized to calculate three kinds of scattering power P of each pixel points，PdAnd Pv：
Wherein, Ps，PdAnd PvSurface scattering power, dihedral scattering power and bulk scattering power are indicated, respectively.
Step 3) respectively calculating the scattering power entropy H of each pixel point by using the three kinds of scattering power of each pixel pointpWherein 0 is not more than Hp≤1；
(3a) Calculating the surface scattering power P of each pixel pointsDihedral angle scattering power PdAnd the bulk scattering power PvThe proportion of the total scattered power is calculated by the following formula:
wherein, P1，P2And P3Are respectively provided withThe proportion of surface scattering power, dihedral angle scattering power and bulk scattering power in the total scattering power is represented;
(3b) calculating the scattering power entropy H of each pixel point in the San Francisco image by using the formula 6)p
Step 4) initial classification of San Francisco images: according to three scattering powers and scattering power entropies H of each pixel point in the San Francisco imagepDividing the San Francisco image into 7 types initially;
the scattering power entropy is a description of the randomness of a scattering mechanism, and the value range is as follows: h is not less than 0pLess than or equal to 1, when HpWhen 0, there is only one scattering mechanism in the data, when H ispWhen 1, the data is an extreme case of a mixture of three scattering mechanisms, when 0 < Hp<1, along with the increase of scattered power entropy, the randomness of scattering process also increases gradually, and when the scattered power entropy is less, only a kind of main scattering mechanism in the data, when the scattered power entropy is great, have two kinds or three kinds of main scattering mechanisms in the data, consequently, the scattered power entropy can be used for confirming the scattering mechanism of pixel, comes to classify the image according to the scattering mechanism of pixel, and the realization step is:
(4a) scattering power entropy H from polarized SAR imagespSelecting two classification thresholds x10.48 and x2＝0.85；
(4b) According to the scattering power entropy HpWith two classification thresholds x1And x2The initial classification of the polarized SAR image is specifically as follows:
when 0 is more than Hp≤x1Will satisfy Ps＝max(Ps,Pd,Pv) The pixel points are classified into surface scattering classes to satisfy Pd＝max(Ps,Pd,Pv) The pixel points are divided into dihedral angle scattering classes satisfying Pv＝max(Ps,Pd,Pv) Is divided into volumesScattering;
when x is1＜Hp≤x2Will satisfy Ps＝min(Ps,Pd,Pv) The pixel points are divided into dihedral angle-volume scattering mixed classes to satisfy Pd＝min(Ps,Pd,Pv) The pixel points are divided into surface-volume scattering mixed classes to satisfy Pv＝min(Ps,Pd,Pv) The pixel points are divided into surface-dihedral angle scattering mixed classes;
when x is2＜HpIf the number of the pixel points is less than or equal to 1, the pixel points are divided into scattering mixed types;
step 5) defining a heteropolarity ratio formula, and calculating the ratio of the co-polarization component to the cross-polarization component of each pixel point in the San Francisco image by using the formula to obtain the heteropolarity ratio of each pixel point;
the heteropolarity ratio represents a ratio between a co-polarized component and a cross-polarized component of the polarized SAR image, and can be a ratio between the co-polarized component and the cross-polarized component or a ratio between the cross-polarized component and the co-polarized component. Neither ratio affects the heteropolarity ratio distribution of the data. The heteropolarity ratio not only contains the co-polarized component of the polarized SAR image, but also contains the cross-polarized component of the image, so that the utilization of information in the polarized SAR image is sufficient, and the classification effect of data can be improved by classifying the image by the heteropolarity ratio. Here, taking the ratio of the common polarization component to the cross polarization component as an example, the calculation formula for defining the heteropolarity ratio is:
wherein SHHRepresenting echo images received horizontally, SVVRepresenting vertically transmitted and vertically received echo images, SHHAnd SVVReferred to as co-polarized component, SHVRepresenting horizontally transmitted and vertically received echo images, SVHRepresenting vertically transmitted horizontally received echo images, SHVAnd SVHReferred to as cross-polarization components, |, denotes takingThe modulus of the number.
Step 6) reclassify the San Francisco images: and subdividing each of the 7 types of San Francisco images in the step 4) in the same proportion according to the heteropolarity ratio of each pixel point in the San Francisco image and the heteropolarity ratio distribution of the San Francisco image. The San Francisco image contains mainly 3 different ground objects: oceans, buildings and vegetation, so we subdivide each of the 7 types of San Francisco images in step 4) into 3 types, resulting in 21 types of San Francisco images, and the implementation steps are as follows:
(6a) two classification thresholds y were selected based on the heteropolarity ratio distribution of the San Francisco image1＝6，y2＝13；
(6b) According to the different polarization ratio R of each pixel in the San Francisco imagedWith two classification thresholds y1And y2In step 4), each of the 7 types of San Francisco images is subdivided into 3 types, specifically: will satisfy Rd≤y1The pixel points of (1) are classified into one class satisfying y1＜Rd≤y2The pixel points are classified into one class satisfying Rd≤y2The pixel points of (a) are classified into one type;
step 7) merging the 21 classes of San Francisco images into 8 classes of San Francisco images by utilizing an inter-class merging criterion based on Wishart distance, wherein the implementation steps are as follows:
(7a) calculating the clustering center of each type of data, wherein the calculation formula is as follows:
wherein i is 1, 2.. 21NiRepresenting the number of pixels in class i data, ClRepresenting a covariance matrix of the ith pixel point in the ith type data;
(7b) calculating the Wishart distance between every two types of data, wherein the calculation formula is as follows:
Dij＝(Ni+Nj)ln|V|-Ni ln|Vi|-Njln|Vj| 9)
wherein i, j is 1, 2.. 21, i ≠ j, and V represents a clustering center after two classes are merged;
(7c) merging the two types of data with the minimum distance;
(7d) and (5) repeating the steps (7a) - (7c) until the total classification number of the polarized SAR images is 8.
Step 8) carrying out complex Wishart iteration on 8 types of data in the San Francisco image, updating the category of each pixel point, and obtaining a more accurate classification result:
(8a) setting the initial iteration time t to be 1 and the maximum iteration time to be 4;
(8b) calculating the clustering center of each type of merged data by using a formula 8);
(8c) calculating the Wishart distance from each pixel to the clustering center of the ith class of data, wherein the calculation formula is as follows:
d(C,Vi)＝ln|Vi|+tr(Vi -1C) 10)
where i is 1, 2.. 8, C is the covariance matrix of the pixel points, tr represents the trace of the matrix, Vi -1Represents ViAn inverse of the matrix;
(8d) reclassifying the pixels according to the Wishart distance from each pixel point to each cluster center: if the pixel satisfies d (C, V)i)≤d(C,Vj) If i, j is 1, 2., 8, i ≠ j, then the pixel is classified as the ith class;
(8e) the iteration number is increased by 1, namely t is t + 1;
(8f) and (5) repeating the steps (8b) to (8e) until the iteration number is 4.
And 9) coloring the classification result obtained in the step 8) according to a three-primary-color principle by using three color components of red R, green G and blue B as three primary colors to obtain a final color classification result image. The coloring principle is generally as follows: the water area part uses blue tone, the artificial building uses red tone, and the vegetation part uses green tone.
The technical effects of the present invention will be further explained below by combining with simulation experiments.
1. Simulation conditions and methods:
simulation environment: intel (R) core (TM) i5-3210M CPU @2.50GHz 2.50GHz Windows 10;
a software platform: matlab2015 a;
the simulation method comprises the following steps: the classification simulation experiment of the invention and the classification method based on the scattering power entropy and the homopolarization ratio on the polarized SAR image is completed.
2. Simulation content and result analysis:
the experimental contents are as follows: the invention uses the San Francisco image shown in FIG. 2 to carry out simulation experiment, the view number of the image is 4, and the size of the image is 900 multiplied by 1024 pixels;
experiment I, the distribution of the heteropolarity ratio of different ground objects in a San Francisco image is analyzed, the San Francisco image mainly comprises three ground object types, such as area 1, area 2 and area 3 marked in figure 2, which respectively represent sea, buildings and vegetation, and the distribution of the heteropolarity ratio of the sea, the buildings and the vegetation is shown in figure 3.
It can be seen from fig. 3 that there is a large difference in the distribution of the heteropolarity ratios between three different surface features, so we can set different heteropolarity ratio classification thresholds to distinguish different surface feature types.
Second, a classification simulation experiment is performed on the San Francisco image by using the classification method based on the scattering power entropy and the homopolarity ratio, and the classification result is shown in fig. 4, wherein fig. 4(a) is the classification result of the classification method based on the scattering power entropy and the homopolarity ratio, and fig. 4(b) is the classification result of the invention.
As can be seen from fig. 4(a), the classification method based on the scattering power entropy and the homopolarity ratio can roughly distinguish different ground object types in the image, but the division of the ocean area is still not fine enough, and a partial region has a wrong division phenomenon.
As can be seen from fig. 4(b), the classification result of the present invention is more accurate, the details of the ground feature are more detailed, and the classification edges of different areas are smoother than those of fig. 4(a), for example, the sea area is more finely divided, the classification result of the building area is closer to the real ground feature, and the consistency of the areas such as the golf course, the horse race and the parking lot is better than that of fig. 4 (a).
In summary, the classification method for the polarized SAR images provided by the invention initially classifies the images by using the three scattering powers and the scattering power entropies of the polarized SAR images, reclassifies the images by using the ratio of the common polarization component to the cross polarization component in the polarized SAR images, and then performs class merging and complex Wishart iteration on the classification results, thereby further improving the classification effect of data, being capable of more accurately distinguishing different ground object types and having better classification performance.

Claims (7)

1. A polarized SAR image classification method based on heteropolarity ratio and scattering power entropy includes the following steps:
(1) removing speckle noise in the polarized SAR image to be classified to obtain a filtered polarized SAR image;
(2) acquiring three scattering powers of each pixel point in the polarized SAR image: performing Freeman decomposition on the filtered polarized SAR image to obtain the surface scattering power P of each pixel pointsDihedral angle scattering power PdAnd bulk scattered power Pv；
(3) Respectively calculating the scattering power entropy H of each pixel point by using the three scattering powers of each pixel pointpWherein 0 is not more than Hp≤1；
(4) Carrying out initial classification on the polarized SAR image: according to three kinds of scattering power and scattering power entropy H of each pixel point in the polarized SAR imagepAnd initially classifying the polarized SAR images to obtain 7 types of polarized SAR images:
(4a) scattering power entropy H from polarized SAR imagespSelecting two classification thresholds x1And x2；
(4b) According to the scattering power entropy HpWith two classification thresholds x1And x2The initial classification of the polarized SAR image is specifically as follows:
when H is more than or equal to 0p≤x1Will satisfy Ps＝max(Ps,Pd,Pv) The pixel points are classified into surface scattering classes to satisfy Pd＝max(Ps,Pd,Pv) The pixel points are divided into dihedral angle scattering classes satisfying Pv＝max(Ps,Pd,Pv) The pixel points are divided into volume scattering classes;
when x is1＜Hp≤x2Will satisfy Ps＝min(Ps,Pd,Pv) The pixel points are divided into dihedral angle-volume scattering mixed classes to satisfy Pd＝min(Ps,Pd,Pv) The pixel points are divided into surface-volume scattering mixed classes to satisfy Pv＝min(Ps,Pd,Pv) The pixel points are divided into surface-dihedral angle scattering mixed classes;
when x is2＜HpIf the number of the pixel points is less than or equal to 1, the pixel points are divided into scattering mixed types;
(5) defining a heteropolarity ratio formula, calculating the ratio of a co-polarized component to a cross-polarized component of each pixel point in the polarized SAR image by using the heteropolarity ratio formula to obtain the heteropolarity ratio of each pixel point, and then counting the heteropolarity ratio distribution of the polarized SAR image;
(6) reclassifying the polarized SAR image: according to the heteropolarity ratio of each pixel point in the polarized SAR image and the heteropolarity ratio distribution of the polarized SAR image, carrying out subdivision on each class in the 7 classes of polarized SAR images in the step (4) in the same proportion to obtain 7n classes of polarized SAR images, wherein n represents the class number of subdivision of each class of data, and n is more than or equal to 2;
(7) performing category merging on the obtained 7n types of polarized SAR images to obtain m types of polarized SAR images, wherein the value of m is determined by the number of surface feature types contained in the polarized SAR images;
(8) updating the category of each pixel point in the m-type polarized SAR images to obtain a classification result;
(9) and (4) coloring the classification result obtained in the step (8) according to a three-primary-color principle by using three color components of red R, green G and blue B as three primary colors to obtain a final color classification result graph.
2. The polarimetric SAR image classification method based on the heteropolarity ratio and the scattering power entropy as claimed in claim 1, characterized in that the speckle noise in the polarimetric SAR image to be classified in step (1) is removed by using a refined Lee filtering method.
3. The polarimetric SAR image classification method based on the heteropolarity ratio and the scattering power entropy as claimed in claim 1, characterized in that, in the step (2), Freeman decomposition is performed on the filtered polarimetric SAR image, and the implementation steps are as follows:
(2a) inputting a covariance matrix C of each pixel point in the polarized SAR image to be classified:
wherein H denotes horizontal polarization, V denotes vertical polarization, SHHCo-polarized component, S, representing horizontal transmission and horizontal receptionVVIndicating the co-polarized component of the vertical transmit vertical receive, SHVRepresents the cross-polarization component of the horizontal transmission and vertical reception, represents the conjugate, | - | represents the modulus,<·>represents a per-view average;
(2b) the covariance matrix C is decomposed into the following form:
wherein f issIs the coefficient of decomposition, f, of the surface scattering componentdIs the decomposition coefficient of the dihedral scattering component, fvBeta represents the ratio of the horizontal emission horizontal reception backscatter reflection coefficient to the vertical emission vertical reception backscatter reflection coefficient, and alpha is defined as alpha-RghRvh/RgvRvv，Rgh，RvhRespectively representing the horizontal and vertical reflection coefficients of the earth's surface, Rgv，RvvRespectively representing the horizontal and vertical reflection coefficients of a vertical wall;
(2c) according to the equality of corresponding elements of the formula 1) and the formula 2), five unknowns f are obtaineds，fd，fvFour equations for α and β:
(2d) solving equation set 3) to obtain fs，fd，fvValues of α and β:
calculating the covariance matrix of each pixelIf the positive and negative values are judged, ifLet α be-1, otherwise let β be 1, solve equation set 3) based on the values of α and β, and then obtain f)s，fdAnd fvWherein Re (. cndot.) represents the real part;
(2e) according to fs，fd，fvAnd the values of alpha and beta, and the following formula is utilized to calculate three kinds of scattering power P of each pixel points，PdAnd Pv：
Wherein, Ps，PdAnd PvSurface scattering power, dihedral scattering power and bulk scattering power are indicated, respectively.
4. The polarimetric SAR image classification method based on the heteropolarity ratio and the scattering power entropy as claimed in claim 1, characterized in that, in the step (3), the scattering power entropy H of each pixel point is calculated by using three kinds of scattering power of each pixel point respectivelypThe method comprises the following implementation steps:
(3a) calculating the surface scattering power P of each pixel pointsDihedral angle scattering power PdAnd the bulk scattering power PvThe proportion of the total scattered power is calculated by the following formula:
wherein, P1，P2And P3Respectively representing the proportion of the surface scattering power, the dihedral angle scattering power and the bulk scattering power in the total scattering power;
(3b) calculating scattering power entropy H of each pixel point in the polarized SAR image by using formula 6)p
。
5. The polarimetric SAR image classification method based on the heteropolarity ratio and the scattering power entropy as claimed in claim 1, characterized in that, in step (5), the ratio of the co-polarized component and the cross-polarized component of each pixel point in the polarimetric SAR image is calculated, and the calculation formula is as follows:
wherein SHHCo-polarized component, S, representing horizontal transmission and horizontal receptionVVIndicating the co-polarized component of the vertical transmit vertical receive, SHVRepresenting the cross-polarized component, S, of the horizontal transmission and vertical receptionVHRepresenting the cross-polarization component of the vertical transmit horizontal reception, |, representing the modulus of the number taken.
6. The polarimetric SAR image classification method based on the heteropolarity ratio and the scattering power entropy as claimed in claim 1, characterized in that, in the step (7), the obtained 7n polarimetric SAR images are subjected to category merging, and the implementation steps are as follows:
(7a) calculating the clustering center of each type of data, wherein the calculation formula is as follows:
wherein i is 1, 2.. 7n NiRepresenting the number of pixels in class i data, ClIndicating the ith in the ith class of dataA covariance matrix of the pixel points;
(7b) calculating the Wishart distance between every two types of data, wherein the calculation formula is as follows:
Dij＝(Ni+Nj)ln|V|-Niln|Vi|-Njln|Vj| 9)
wherein i, j is 1,2,. 7n, i ≠ j, and V represents a clustering center after two classes are merged;
(7c) merging the two types of data with the minimum distance;
(7d) and (5) repeating the steps (7a) - (7c) until the total classification number of the polarized SAR images is m.
7. The polarimetric SAR image classification method based on the heteropolarity ratio and the scattering power entropy as claimed in claim 6, characterized in that, in step (8), the category to which each pixel point in the m polarimetric SAR images belongs is updated, and the implementation steps are as follows:
(8a) setting the initial iteration time T to be 1, and setting the maximum iteration time to be T, wherein T is more than 1;
(8b) calculating the clustering center of each type of merged data by using a formula 8);
(8c) calculating the Wishart distance from each pixel to the clustering center of the ith class of data, wherein the calculation formula is as follows:
d(C,Vi)＝ln|Vi|+tr(Vi -1C) 10)
where i ═ 1, 2.. m, C is the covariance matrix of the pixel points, tr denotes the trace of the matrix, Vi -1Represents ViAn inverse of the matrix;
(8d) reclassifying the pixels according to the Wishart distance from each pixel point to each cluster center: if the pixel satisfies d (C, V)i)≤d(C,Vj) If i, j is 1, 2., m, i ≠ j, the pixel is classified as the ith class;
(8e) the iteration number is increased by 1, namely t is t + 1;
(8f) and (4) repeating the steps (8b) - (8e) until the maximum iteration number is reached, namely T ═ T is met, wherein T is larger than 1.
CN201710522622.4A 2017-06-30 2017-06-30 Polarimetric SAR image classification method based on heteropolarity ratio and scattering power entropy Active CN107203791B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN201710522622.4A CN107203791B (en) 2017-06-30 2017-06-30 Polarimetric SAR image classification method based on heteropolarity ratio and scattering power entropy

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN201710522622.4A CN107203791B (en) 2017-06-30 2017-06-30 Polarimetric SAR image classification method based on heteropolarity ratio and scattering power entropy

Publications (2)

Publication Number Publication Date
CN107203791A CN107203791A (en) 2017-09-26
CN107203791B true CN107203791B (en) 2019-12-31

Family

ID=59910448

Family Applications (1)

Application Number Title Priority Date Filing Date
CN201710522622.4A Active CN107203791B (en) 2017-06-30 2017-06-30 Polarimetric SAR image classification method based on heteropolarity ratio and scattering power entropy

Country Status (1)

Country Link
CN (1) CN107203791B (en)

Families Citing this family (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN108872987A (en) * 2018-06-29 2018-11-23 中国科学院电子学研究所 The target extraction method with cylindrical structure based on polarization circular track SAR data
CN110646795B (en) * 2019-09-16 2021-09-24 武汉大学 Ocean oil spill detection method and system of simple polarization SAR
CN112529817B (en) * 2020-12-23 2022-11-08 中国科学院空天信息创新研究院 Sea spike suppression method for uncalibrated airborne SAR image based on optimal polarization ratio
CN115063687A (en) * 2022-08-19 2022-09-16 航天宏图信息技术股份有限公司 Polarized SAR image classification method and device
CN116797845B (en) * 2023-07-05 2024-01-26 中国科学院空天信息创新研究院 Unsupervised reduced polarization classification method based on scattering mechanism

Citations (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN103413146A (en) * 2013-08-23 2013-11-27 西安电子科技大学 Method for finely classifying polarized SAR images based on Freeman entropy and self-learning
CN104504393A (en) * 2014-12-04 2015-04-08 西安电子科技大学 SAR (Synthetic Aperture Radar) image semi-supervised classification method based on integrated learning
CN105160353A (en) * 2015-08-18 2015-12-16 西安电子科技大学 Polarimetric SAR data ground object classification method based on multiple feature sets

Family Cites Families (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US8620093B2 (en) * 2010-03-15 2013-12-31 The United States Of America As Represented By The Secretary Of The Army Method and system for image registration and change detection
US10235589B2 (en) * 2015-09-08 2019-03-19 The Johns Hopkins University Small maritime target detector

Patent Citations (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN103413146A (en) * 2013-08-23 2013-11-27 西安电子科技大学 Method for finely classifying polarized SAR images based on Freeman entropy and self-learning
CN104504393A (en) * 2014-12-04 2015-04-08 西安电子科技大学 SAR (Synthetic Aperture Radar) image semi-supervised classification method based on integrated learning
CN105160353A (en) * 2015-08-18 2015-12-16 西安电子科技大学 Polarimetric SAR data ground object classification method based on multiple feature sets

Non-Patent Citations (2)

* Cited by examiner, † Cited by third party
Title
《Unsupervised classification of fully polarimetric SAR images based on scattering power entropy and copolarized ratio》;WANG Shuang，et al.;《IEEE Geoscience and Remote Sensing Letters》;20131231;第10卷(第3期);第622-626页 *
《一种新的极化SAR图像非监督分类算法研究》;邹同元，et al.;《武汉大学学报 信息科技版》;20090831;第34卷(第8期);第910-914页 *

Also Published As

Publication number Publication date
CN107203791A (en) 2017-09-26

Similar Documents

Publication Publication Date Title
CN107203791B (en) Polarimetric SAR image classification method based on heteropolarity ratio and scattering power entropy
Wang et al. Unsupervised classification of fully polarimetric SAR images based on scattering power entropy and copolarized ratio
CN102208031B (en) Freeman decomposition and homo-polarization rate-based polarized synthetic aperture radar (SAR) image classification method
Mishra et al. Land cover classification of PALSAR images by knowledge based decision tree classifier and supervised classifiers based on SAR observables
CN103247059B (en) A kind of remote sensing images region of interest detection method based on integer wavelet and visual signature
CN102968640B (en) Decompose and the Classification of Polarimetric SAR Image method of data distribution characteristics based on Freeman
CN103839073B (en) Polarization SAR image classification method based on polarization features and affinity propagation clustering
CN102999762B (en) Decompose and the Classification of Polarimetric SAR Image method of spectral clustering based on Freeman
Anfinsen Statistical analysis of multilook polarimetric radar images with the Mellin transform
CN105069796B (en) SAR image segmentation method based on small echo both scatternets
CN105138966B (en) Classification of Polarimetric SAR Image method based on fast density peak value cluster
CN105160353B (en) Polarization SAR data terrain classification method based on multiple features collection
CN109840542B (en) Self-adaptive dimension decision tree classification method based on polarization characteristics
CN103186794A (en) Polarized SAT (synthetic aperture radar) image classification method based on improved affinity propagation clustering
CN103903012A (en) Polarimetric SAR data classifying method based on orientation object and support vector machine
CN104700116B (en) The sorting technique of the Polarimetric SAR Image atural object represented based on multi-layer quantum ridge ripple
CN102999761A (en) Method for classifying polarimetric SAR (synthetic aperture radar) images on the basis of Cloude decomposition and K-wishart distribution
CN103365985B (en) The adaptive polarization SAR sorting technique of one kind
CN111626380A (en) Polarized SAR image classification method based on super-pixels and convolution network
CN112883877B (en) Method and system for extracting polarized SAR mangrove based on double unsupervised classification
Weissgerber et al. A temporal estimation of entropy and its comparison with spatial estimations on PolSAR images
CN103886327B (en) Polarized SAR image classification method based on 2D-KPCA
Kourgli et al. Land cover identification using polarimetric SAR images
CN103870841A (en) Polarization SAR (synthetic aperture radar) image classification method based on Freeman decomposition and PSO (particle swarm optimization)
Liu et al. Unsupervised classification of polarimetric SAR images integrating color features

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