CN114494282A - Complex background downslope identification method and device fusing polarization information - Google Patents

Complex background downslope identification method and device fusing polarization information Download PDF

Info

Publication number
CN114494282A
CN114494282A CN202111598126.XA CN202111598126A CN114494282A CN 114494282 A CN114494282 A CN 114494282A CN 202111598126 A CN202111598126 A CN 202111598126A CN 114494282 A CN114494282 A CN 114494282A
Authority
CN
China
Prior art keywords
landslide
polarization
scattering
area
image
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
CN202111598126.XA
Other languages
Chinese (zh)
Inventor
牛朝阳
高欧阳
刘伟
李润生
卢万杰
邹玮琦
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.)
Information Engineering University of PLA Strategic Support Force
Original Assignee
Information Engineering University of PLA Strategic Support Force
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 Information Engineering University of PLA Strategic Support Force filed Critical Information Engineering University of PLA Strategic Support Force
Priority to CN202111598126.XA priority Critical patent/CN114494282A/en
Publication of CN114494282A publication Critical patent/CN114494282A/en
Pending legal-status Critical Current

Links

Images

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T7/00—Image analysis
    • G06T7/10—Segmentation; Edge detection
    • G06T7/11—Region-based segmentation
    • G—PHYSICS
    • G01—MEASURING; TESTING
    • G01S—RADIO DIRECTION-FINDING; RADIO NAVIGATION; DETERMINING DISTANCE OR VELOCITY BY USE OF RADIO WAVES; LOCATING OR PRESENCE-DETECTING BY USE OF THE REFLECTION OR RERADIATION OF RADIO WAVES; ANALOGOUS ARRANGEMENTS USING OTHER WAVES
    • G01S13/00—Systems using the reflection or reradiation of radio waves, e.g. radar systems; Analogous systems using reflection or reradiation of waves whose nature or wavelength is irrelevant or unspecified
    • G01S13/88—Radar or analogous systems specially adapted for specific applications
    • G01S13/89—Radar or analogous systems specially adapted for specific applications for mapping or imaging
    • G01S13/90—Radar or analogous systems specially adapted for specific applications for mapping or imaging using synthetic aperture techniques, e.g. synthetic aperture radar [SAR] techniques
    • G—PHYSICS
    • G01—MEASURING; TESTING
    • G01S—RADIO DIRECTION-FINDING; RADIO NAVIGATION; DETERMINING DISTANCE OR VELOCITY BY USE OF RADIO WAVES; LOCATING OR PRESENCE-DETECTING BY USE OF THE REFLECTION OR RERADIATION OF RADIO WAVES; ANALOGOUS ARRANGEMENTS USING OTHER WAVES
    • G01S13/00—Systems using the reflection or reradiation of radio waves, e.g. radar systems; Analogous systems using reflection or reradiation of waves whose nature or wavelength is irrelevant or unspecified
    • G01S13/88—Radar or analogous systems specially adapted for specific applications
    • G01S13/89—Radar or analogous systems specially adapted for specific applications for mapping or imaging
    • G01S13/90—Radar or analogous systems specially adapted for specific applications for mapping or imaging using synthetic aperture techniques, e.g. synthetic aperture radar [SAR] techniques
    • G01S13/9021—SAR image post-processing techniques
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T7/00—Image analysis
    • G06T7/10—Segmentation; Edge detection
    • G06T7/136—Segmentation; Edge detection involving thresholding
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T7/00—Image analysis
    • G06T7/30—Determination of transform parameters for the alignment of images, i.e. image registration
    • G06T7/33—Determination of transform parameters for the alignment of images, i.e. image registration using feature-based methods

Landscapes

  • Engineering & Computer Science (AREA)
  • Remote Sensing (AREA)
  • Radar, Positioning & Navigation (AREA)
  • Physics & Mathematics (AREA)
  • General Physics & Mathematics (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • Theoretical Computer Science (AREA)
  • Electromagnetism (AREA)
  • Computer Networks & Wireless Communication (AREA)
  • Image Analysis (AREA)
  • Pit Excavations, Shoring, Fill Or Stabilisation Of Slopes (AREA)

Abstract

The invention belongs to the technical field of image processing of a fully-polarized synthetic aperture radar, and particularly relates to a method and a device for identifying a slope with a complex background fused with polarization information, wherein the method comprises the steps of registering a unipolar SAR image before the slope and an image in a polar SAR image after the slope, which has the same polarization combination mode as the unipolar SAR image before the slope, constructing a double-time-phase SAR image pair, and generating a coherent graph through coherent change detection to extract a change area; extracting multiple polarization characteristic parameters from the preprocessed landslide PolSAR image, and fully fusing the polarization characteristic parameters by using TOPSIS with weight to perform landslide detection to obtain a suspected landslide area; and fusing the change area and the suspected landslide area extracted from the coherent graph through logical AND operation to obtain a final landslide area. The landslide detection method based on the TOPSIS fusion polarization information utilizes the TOPSIS fusion polarization information to carry out landslide detection, and combines with coherent change detection, so that the landslide detection accuracy is greatly improved under the condition of effectively inhibiting the false alarm rate.

Description

Complex background downslope identification method and device fusing polarization information
Technical Field
The invention belongs to the technical field of image processing of a polarimetric synthetic aperture radar (PolSAR), and particularly relates to a method and a device for identifying a slope under a complex background by fusing polarization information.
Background
Landslide is one of natural disasters which frequently occur and seriously harm, can directly cause damage to buildings and interruption of transportation, and brings huge loss to lives and properties of residents. The method utilizes a Synthetic Aperture Radar (SAR) remote sensing satellite image to detect the position and the boundary of the landslide after the disaster, investigates the spatial distribution characteristics of the landslide and can provide important information for disaster prevention and control and disaster area reconstruction. Landslide detection is carried out on the basis of polarization decomposition, wherein the landslide detection is carried out by utilizing a multi-threshold method, namely thresholds are respectively set for multiple polarization characteristics, regions meeting all threshold conditions are determined as landslide regions, and post-disaster PolSAR images are classified by utilizing a supervised classification or unsupervised classification method to distinguish the regions belonging to landslides. Compared with the multi-threshold method, the latter method can provide higher landslide detection accuracy due to the complex structure of the classifier, so that the classification method is more commonly used for landslide identification than the multi-threshold method. Such methods, however, do not effectively distinguish regions that resemble the scattering characteristics of a landslide. In recent years, researchers have combined Change Detection (CD) with Analytic Hierarchy Process (AHP) to achieve landslide Detection in complex topographical backgrounds. In the method, CD only utilizes the scattering information of the PolSAR image, and does not utilize phase information, so that false detection can be generated in a region with a similar scattering mechanism; in the process of fusing the polarization information by the AHP, the method has certain subjectivity due to the use of an expert scoring method. Therefore, a method for accurately identifying a landslide area under a complex background with a plurality of surface feature types and better inhibiting false alarms caused by other surface features needs to be provided.
Disclosure of Invention
Aiming at the problems in the prior art, the invention provides a method and a device for identifying the landslide of the complex background fused with polarization information, which are used for carrying out landslide detection by fusing the polarization information through a superior-inferior solution distance method (TOPSIS), and greatly improving the landslide detection accuracy under the condition of effectively inhibiting the false alarm rate by combining with coherent change detection.
In order to achieve the purpose, the invention adopts the following technical scheme:
the invention provides a method for identifying a complex background downslide by fusing polarization information, which comprises the following steps:
registering images, which have the same polarization combination mode as the pre-landslide single-polarized SAR image, in the pre-landslide single-polarized SAR image and the post-landslide PolSAR image, constructing a double-time-phase SAR image pair, and generating a coherent image through coherent change detection to extract a change area;
sequentially carrying out radiometric calibration, multi-view and terrain correction preprocessing on the PolSAR image after landslide; extracting multiple polarization characteristic parameters from the preprocessed landslide PolSAR image, and fully fusing the polarization characteristic parameters by using weighted TOPSIS to perform landslide detection to obtain a region with the same scattering characteristics as the landslide surface, namely a suspected landslide region;
and fusing the change area extracted from the coherent graph and the suspected landslide area through logical AND operation, and inhibiting false alarm brought by ground objects with similar scattering characteristics to the real landslide area in the suspected landslide area through the change area extracted from the coherent graph to obtain a final landslide area.
Further, a coherence map is obtained by using a bi-temporal SAR image pair, wherein the size of a gray value in the coherence map indicates the level of coherence, the abrupt nature of a landslide can cause a terrain landform to change significantly, namely the coherence is low, an unchanged area maintains high coherence, and the coherence | γ | is calculated as follows:
Figure BDA0003430966630000021
wherein, E2]Representing a mathematical periodInspection of S1Is SAR image before landslide, S2The image which is selected from the PolSAR image after the landslide and has the same polarization combination mode as the single-polarized SAR image before the landslide,
Figure BDA0003430966630000031
denotes S2The value range of the coherence | gamma | is [0,1 ]]The larger the | gamma | is, the lower the change degree of a region is, and the smaller the | gamma | is, the more obvious the change of the region is; thresholding and morphological processing of the coherence map using coherence | γ | yields regions of significant variation.
Further, the surface scattering power p is obtained by Yamaguchi decompositionsSecondary scattering power pdVolume scattering power pvSum helical scattering power ph(ii) a Obtaining polarization entropy H, average scattering angle alpha and inverse entropy A through H/A/alpha decomposition; calculating the real part 0e (rho) of the correlation coefficient of the same polarization component through the correlation coefficienthh-vv)。
Further, extracting the power corresponding to each scattering component from the polarized coherent matrix T through Yamaguchi decomposition, namely surface scattering power PsSecondary scattering power PdVolume scattering power PvSum helical scattering power Ph(ii) a The polarization coherence matrix T is represented as a weighted combination of the individual scattered powers:
T＝fsTs+fdTd+fvTv+fhTh
wherein, Ts、Td、Tv、ΤhPolarization coherent matrixes corresponding to surface scattering, secondary scattering, volume scattering and spiral body scattering components are respectively adopted; f. ofs、fd、fv、fhAre respectively Ps、Pd、PvAnd PhThe decomposition coefficient of (a).
Further, polarization entropy H, average scattering angle alpha and inverse entropy A are obtained through H/A/alpha decomposition, and the method comprises the following steps: under the condition of single-station reciprocity, the polarization coherent matrix T is a semi-positive definite Hermite matrix of 3 multiplied by 3, so the following eigenvalue decomposition is carried out:
Figure BDA0003430966630000032
wherein the characteristic value lambdaiI is more than or equal to 1 and less than or equal to 3, and satisfies lambda1≥λ2≥λ3>0，λiCorresponding feature vector uiMutually orthogonal, H represents conjugate transpose; the polarization entropy H, the mean scattering angle alpha and the inverse entropy A are determined by the eigenvalues lambdaiAnd a feature vector uiAnd calculating to obtain:
Figure BDA0003430966630000041
in the formula (I), the compound is shown in the specification,
Figure BDA0003430966630000042
for the probability of occurrence of each scattering mechanism,
Figure BDA0003430966630000043
is a feature vector uiThe corresponding scattering angle.
Further, the co-polarized component correlation coefficient ρhh-vvCalculating Re (rho) by a polarization coherence matrix T for a correlation coefficient between two homopolarization scattering vectors of the target, wherein the larger the value is, the more single the scattering type of the target is shown, andhh-vv) The formula of (1) is as follows:
Figure BDA0003430966630000044
wherein Re (. cndot.) represents the real part, T11、T12And T22Representing the corresponding elements in the polarization coherence matrix T.
Further, the landslide detection by using the weighted TOPSIS fully fused polarization characteristic parameter comprises the following steps:
firstly, carrying out normalization and homotrending processing on polarization characteristic parameters;
taking the eight polarization characteristic parameters as evaluation indexes, and taking each pixel point of the PolSAR image as an evaluation object to obtain an evaluation matrix;
determining that a positive ideal sample of the evaluation matrix is 1 and a negative ideal sample of the evaluation matrix is 0;
calculating the distance between each pixel point and the positive and negative ideal samples;
calculating the relative closeness of each pixel point to a positive ideal sample;
and determining a proper threshold value to perform threshold segmentation on the relative closeness of each pixel point of the PolSAR image, determining the pixel points with the relative closeness larger than the threshold value as suspected landslide pixel points, and performing morphological processing to obtain a suspected landslide area.
Further, the normalizing and homotrending processing of the polarization characteristic parameters includes:
for the scattered power psSecondary scattering power pdVolume scattering power pvSum helical scattering power phAnd (3) carrying out normalization treatment:
Figure BDA0003430966630000051
wherein, the subscript x belongs to { s, d, v, h }, and s, d, v and h respectively represent surface scattering, secondary scattering, volume scattering and spirochete scattering; p after normalization processingd,pv,phCarrying out forward homotrend processing:
px＝1-px
and (3) carrying out interval type normalization processing on the polarization entropy H and the average scattering angle alpha:
Figure BDA0003430966630000052
wherein, for the H, the first and second groups are,
Figure BDA0003430966630000053
in the case of a, for a,
Figure BDA0003430966630000054
real part Re (rho) of the correlation coefficient for the same polarization componenthh-vv) Re (ρ)hh-vv) The negative part of (a) is assigned 0.
Further, the weight of the polarization characteristic parameter is determined by using an AHP method, and p is the landform characteristic after landslide disastersThe surface scattering characteristic of the landslide can be reflected to the greatest extent, H and alpha can better distinguish the landslide from other landforms, Re (rho)hh-vv) Surface scattering properties, p, commonly used to reflect topographyd、pv、phAnd A has little effect on landslide identification; thus p will besH and alpha are the most important evaluation indexes, and Re (rho)hh-vv) As a generally important evaluation index, pd、pv、phAnd A as a relatively unimportant evaluation index; and then determining importance comparison values among different evaluation indexes according to the pair comparison matrix importance scale table to construct a judgment matrix.
The invention also provides a device for identifying the complex background downslope by fusing the polarization information, which comprises the following components:
the change detection module is used for registering images, which have the same polarization combination mode with the pre-landslide single-polarized SAR image, in the pre-landslide single-polarized SAR image and the post-landslide PolSAR image, constructing a double-time phase SAR image pair, and generating a coherent graph through coherent change detection to extract a change area;
the suspected landslide area extraction module is used for sequentially carrying out radiometric calibration, multi-view and terrain correction pretreatment on the post-landslide PolSAR image; extracting multiple polarization characteristic parameters from the preprocessed landslide PolSAR image, and fully fusing the polarization characteristic parameters by using weighted TOPSIS to perform landslide detection to obtain a region with the same scattering characteristics as the landslide surface, namely a suspected landslide region;
and the landslide area determination module is used for fusing the change area extracted from the coherent graph and the suspected landslide area through logical AND operation, and inhibiting false alarm brought by surface features similar to scattering characteristics of the real landslide area in the suspected landslide area through the change area extracted from the coherent graph to obtain a final landslide area.
Compared with the prior art, the invention has the following advantages:
1. the invention provides a complex background downslide slope identification method fusing polarization information, aiming at the problems of low accuracy and high false alarm rate of landslide detection in a complex background in the prior art, and the method comprises the steps of firstly obtaining a change area through coherent change detection, then utilizing TOPSIS to fuse the polarization information, detecting a suspected landslide area on the basis of scattering characteristics of different landforms by integrating a plurality of polarization characteristic parameters, further fusing the change area and the suspected landslide area, extracting a final landslide area from the change area, obtaining a clear landslide area range, greatly improving the landslide detection accuracy, and effectively inhibiting false alarms caused by most landforms.
2. With the continuous development of the PolSAR technology, the method for disaster detection by utilizing the PolSAR image plays an increasingly important role, can accurately and quickly extract information of a disaster area after landslide without relying on prior information, provides a specific feasible scheme for landslide detection under a complex background, and provides a reliable basis for activities such as post-disaster rescue, secondary disaster early warning and post-disaster reconstruction.
Drawings
In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings used in the embodiments or the prior art descriptions will be briefly described below, it is obvious that the drawings in the following description are only some embodiments of the present invention, and other drawings can be obtained by those skilled in the art without creative efforts.
Fig. 1 is a schematic flowchart of a slope-descending identification method with a complex background fused with polarization information according to an embodiment of the present invention;
FIG. 2 is a schematic flow chart of determining a suspected landslide area using TOPSIS in accordance with an embodiment of the present invention;
fig. 3 is a Yamaguchi decomposed false color image and an optical image before and after occurrence of a landslide according to an embodiment of the present invention, in which (a) is the Yamaguchi decomposed false color image before landslide, (b) is the Yamaguchi decomposed false color image after landslide, (c) is the optical image before landslide, and (d) is the optical image after landslide;
FIG. 4 is a graph comparing visual interpretation results of embodiments of the present invention and methods of the present invention.
Detailed Description
The technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the drawings in the embodiments of the present invention, and it is obvious that the described embodiments are only a part of the embodiments of the present invention, and not all of the embodiments. All other embodiments, which can be derived by a person skilled in the art from the embodiments given herein without making any creative effort, shall fall within the protection scope of the present invention.
How to accurately identify a landslide area under a complex background with various ground object types and better inhibit false alarms caused by other ground objects is a key problem of landslide detection based on a polarized synthetic aperture radar (PolSAR) image. To solve this problem, as shown in fig. 1, this embodiment proposes a complex background downslope identification method fusing polarization information, introduces the ideas of multiple polarization feature fusion and coherent change detection, and includes the following steps:
and step S11, registering images in the pre-landslide single-polarized SAR image and the post-landslide PolSAR image which have the same polarization combination mode as the pre-landslide single-polarized SAR image, constructing a double-time phase SAR image pair, and generating a coherence map through coherent change detection to extract a change region.
Step S12, sequentially carrying out radiometric calibration, multi-view and terrain correction preprocessing on the PolSAR image after landslide; extracting multiple polarization characteristic parameters from the preprocessed landslide PolSAR image, and fully fusing the polarization characteristic parameters by using weighted TOPSIS to perform landslide detection to obtain a region with the same scattering characteristics as the landslide surface, namely a suspected landslide region.
And step S13, fusing the change area and the suspected landslide area extracted from the coherent graph through logical AND operation, and inhibiting false alarms brought by ground objects with scattering characteristics similar to those of the real landslide area in the suspected landslide area through the change area extracted from the coherent graph to obtain a final landslide area so as to achieve the purposes of accurately identifying the landslide area under a complex background with various ground object types and better inhibiting the false alarms caused by other ground objects.
Landslide is taken as a sudden natural disaster, a disaster area has very remarkable mutability, and unchanged land feature and landform generally do not have the characteristics, so that the changed area can be detected through mutational analysis of the land feature and the landform. The obvious change of the landform usually causes serious phase loss coherence in SAR interference processing, namely the coherence is low, and the unchanged area keeps high coherence, so that whether a certain area changes or not can be judged through a coherence map, and then a false alarm caused by unchanged ground objects is inhibited, and the landslide detection result is more accurate.
Obtaining a coherence map by using the bi-temporal SAR image pair, wherein the size of a gray value in the coherence map represents the height of coherence, and the coherence | gamma | is calculated as follows:
Figure BDA0003430966630000081
wherein, E2]Representing a mathematical expectation, S1Is SAR image before landslide, S2The image which is selected from the PolSAR image after the landslide and has the same polarization combination mode as the single-polarized SAR image before the landslide,
Figure BDA0003430966630000082
denotes S2The value range of the coherence | gamma | is [0,1 ]]The larger the | gamma | is, the lower the change degree of a region is, and the smaller the | gamma | is, the more obvious the change of the region is; thresholding and morphological processing of the coherence map using coherence | γ | yields regions of significant variation.
In SAR images with different polarization combinations, vertical polarization backscattering can better reflect landforms mainly with horizontal structures, so that the example adopts a VV polarization combination image capable of highlighting the characteristics of bare soil and bare rock areas to generate a coherence map of a research area.
Specifically, the surface scattering power p is obtained by Yamaguchi decompositionsSecondary scattering power pdVolume scattering power pvSum helical scattering power ph(ii) a Obtaining polarization entropy H, average scattering angle alpha and inverse entropy A through H/A/alpha decomposition; calculating the real part Re (rho) of the correlation coefficient of the same polarization component through the correlation coefficienthh-vv)。
(1) Yamaguchi decomposition
Yamaguchi decomposition is an incoherent decomposition algorithm based on a scattering model, and can well match with a basic scattering mechanism of ground objects. The decomposition method extracts the power corresponding to each scattering component from the polarization coherent matrix T, namely the surface scattering power PsSecondary scattering power PdVolume scattering power PvSum helical scattering power Ph(ii) a The polarization coherence matrix T can be expressed as a weighted combination of the individual scattered powers:
T＝fsTs+fdTd+fvTv+fhTh
wherein, Ts、Td、Tv、ΤhPolarization coherent matrixes corresponding to surface scattering, secondary scattering, volume scattering and spiral body scattering components are respectively adopted; f. ofs、fd、fv、fhAre respectively Ps、Pd、PvAnd PhThe decomposition coefficient of (a).
(2) H/A/alpha decomposition
The H/a/α decomposition is a non-coherent decomposition algorithm that defines polarization features using eigenvalues and eigenvectors of a covariance matrix or a polarization coherence matrix. Under the condition of single-station reciprocity, the polarization coherent matrix T is a semi-positive definite Hermite matrix of 3 multiplied by 3, so the following eigenvalue decomposition is carried out:
Figure BDA0003430966630000091
wherein the characteristic valueλi(1. ltoreq. i. ltoreq.3) satisfies lambda1≥λ2≥λ3>0，λiCorresponding feature vector uiMutually orthogonal, H represents conjugate transpose; the polarization entropy H, the mean scattering angle alpha and the inverse entropy A are determined by the eigenvalues lambdaiAnd a feature vector uiAnd calculating to obtain:
Figure BDA0003430966630000101
in the formula (I), the compound is shown in the specification,
Figure BDA0003430966630000102
for the probability of occurrence of each scattering mechanism,
Figure BDA0003430966630000103
is a feature vector uiThe corresponding scattering angle.
(3) Correlation coefficient calculation
Correlation coefficient rho of homopolarized componenthh-vvFor the correlation coefficient between two homopolarity scattering vectors of the target, a larger value indicates a more unique scattering type of the target, such as Re (rho)hh-vv) The positive portion of (a) is often used to reflect the surface scattering properties of the feature. Calculation of Re (ρ) by polarizing the coherence matrix Thh-vv) The formula of (1) is as follows:
Figure BDA0003430966630000104
wherein Re (. cndot.) represents the real part, T11、T12And T22Representing the corresponding elements in the polarization coherence matrix T.
TOPSIS belongs to a common method in multi-attribute decision analysis, also called an approximation ideal method, has no strict limitation on data distribution and sample content, can fully utilize original data information, and has a more accurate and reliable evaluation result. The method ranks the closeness of a limited number of evaluation objects to an ideal sample according to a series of attribute conditions (namely, evaluation indexes). In step S12, the weighted TOPSIS is used to sufficiently fuse the polarization characteristic parameters for landslide detection, as shown in fig. 2, which includes the following steps:
and step S121, carrying out normalization and homotrending processing on the polarization characteristic parameters.
In step S122, it is assumed that N evaluation objects are evaluated using M evaluation indexes, where z is the value of the ith (i ═ 1,2,3, …, N) evaluation index of the ith (j ═ 1,2,3, …, M) evaluation objectNMThe evaluation matrix can be derived as follows:
Figure BDA0003430966630000111
in the embodiment, each pixel point of the PolSAR image is used as an evaluation object, each polarization characteristic parameter is used as an evaluation index of the score of the evaluation object, and each pixel point of the PolSAR image is evaluated respectively by adopting the eight polarization characteristic parameters.
Step S123, defining the positive and negative ideal samples of the evaluation matrix as:
Figure BDA0003430966630000112
Figure BDA0003430966630000113
to ensure applicability of toposis to different PolSAR image data, a positive ideal sample is determined to be 1 and a negative ideal sample is determined to be 0.
In step S124, the distances between the ith evaluation object (i.e. the ith pixel point) and the positive and negative ideal samples are respectively:
Figure BDA0003430966630000114
Figure BDA0003430966630000115
in the formula wiThe weight representing each evaluation index (i.e., each polarization characteristic parameter) may be determined by different methods according to different requirements.
Step S125, calculating a final score of the evaluation object, that is, a relative closeness to the ideal sample is:
Figure BDA0003430966630000116
relative closeness SiThe size of (A) reflects the difference between the evaluation object and the ideal sample, SiThe larger the evaluation object, the closer the evaluation object is to the positive ideal sample, SiSmaller means that the evaluation object is farther from the positive ideal sample; and accurately calculating the similarity, namely the relative closeness, of each pixel point of the PolSAR image and the landslide through the formula.
Step S126, a proper threshold is determined to carry out threshold segmentation on the relative closeness of each pixel point of the PolSAR image, the pixel points with the relative closeness larger than the threshold are determined to be suspected landslide pixel points, and then morphological processing is carried out to obtain a suspected landslide area.
Morphological treatment: the binary image of the suspected landslide area is processed by adopting the operation of opening firstly and closing secondly, the interference of irrelevant structures in the image is eliminated, white spots (such as salt and pepper noise) in a black background are eliminated through the opening operation firstly, then black holes in a white area are eliminated through the closing operation, the adjacent white area is connected, and the boundary of the white area is smoothed, so that the boundary characteristic of the suspected landslide area is more obvious.
Because of eight polarization characteristic parameters (p)s、pd、pv、ph、H、A、α、Re(ρhh-vv) Differences in directivity, metric unit, numerical evaluation method, and evaluation value range as evaluation indexes adversely affect the evaluation results, and therefore normalization and homotrenization of these polarization characteristic parameters are required before the evaluation matrix is constructed so that the values of each parameter are in the interval [0,1 ]]Internal and positively correlated with landslide scattering propertiesThe larger the value, the more the scattering characteristic of the landslide is exhibited. The normalization and homotrending process of the polarization characteristic parameters in step S121 includes:
(1) for the scattered power psSecondary scattering power pdVolume scattering power pvSum helical scattering power phAnd (3) carrying out normalization treatment:
Figure BDA0003430966630000121
wherein, the subscript x belongs to { s, d, v, h }, and s, d, v and h respectively represent surface scattering, secondary scattering, volume scattering and spirochete scattering; since the bare soil covered by the earth's surface after the occurrence of a landslide is usually present in the form of surface scattering, which is very different from other ground features, psThe larger the value, pd,pv,phThe smaller the value of (A), the more the surface scattering type of landslide landform can be reflected, so that the normalized p isd,pv,phCarrying out forward homotrend processing:
px＝1-px
(2) for polarization entropy H and average scattering angle alpha, when H is 0.52-0.63 and alpha is in a value range of 29-37, the scattering characteristic of landslide can be reflected, and interval type normalization processing is carried out on the polarization entropy H and the average scattering angle alpha:
Figure BDA0003430966630000131
wherein, for the H, the first and second groups are,
Figure BDA0003430966630000132
in the case of a, for a,
Figure BDA0003430966630000133
(3)Re(ρhh-vv) The positive value part of (2) is used for identifying the surface scattering ground object type, and the larger the value is, the more obvious the surface scattering ground object type isRe (ρ)hh-vv) The negative part of (a) is assigned 0.
W in step S124iThe weight of each polarization characteristic parameter is expressed, and different polarization characteristic parameters have different influences on final evaluation, so that different weights need to be given to the polarization characteristic parameters. According to the features of landslide after disaster, psThe surface scattering characteristic of the landslide can be reflected most, H and alpha can better distinguish the landslide from other landforms, namely Re (rho)hh-vv) Surface scattering properties, p, commonly used to reflect topographyd、pv、phAnd A has little effect on landslide identification; thus p will besH and alpha are the most important evaluation indexes, and Re (rho)hh-vv) As a generally important evaluation index, pd、pv、phAnd A as a relatively unimportant evaluation index; then, the importance comparison values among different evaluation indexes are determined according to the pair comparison matrix importance scale table, and a judgment matrix is constructed as shown in table 1. In order to eliminate the subjectivity existing in the judgment matrix as much as possible, a Consistency Ratio (CR) is required to be introduced as a reference when the CR is used<The judgment matrix at 0.1 is considered reasonable, and the judgment matrix CR shown in table 1 is calculated to be 0, which indicates that the weight assignment of each polarization characteristic parameter is reasonable.
TABLE 1 determination matrix for determining weights by AHP method
Figure BDA0003430966630000141
A specific experiment is given below to better illustrate the present invention.
(1) Study area and data source
The study area was: the landslide caused by rainstorm in 12 th and 20 th of 2015 occurs in the Shenzhen Hengtaiyu industrial park (north latitude 22.7065-22.7289 degrees, east longitude 113.9220-113.9467 degrees). The landslide occurs in a complex topographic background (including landslide, water system, forest, residential land and field five basic land features), and other changes caused by natural or human activities exist, so that the area is used as a research object to have certain representativeness, and satellite data are shown in a table 2.
TABLE 2 basic parameters of satellite data
Figure BDA0003430966630000142
The Yamaguchi resolved false color image and optical image of the study area is shown in fig. 3, where red for the false color image represents secondary scattering, green for volume scattering, and blue for surface scattering. There is a certain difference between the acquisition time of the optical image before and after the disaster and the acquisition time of the PolSAR image, but the PolSAR image can still be used as a reference for analyzing and discussing the change of landform caused by landslide. After visual interpretation, five types of ground object samples of landslide, water system, forest, residential land and field are selected on the optical image and the Yamaguchi exploded view by using a white frame, so that detection accuracy analysis can be conveniently made below.
(2) Quantitative analysis of landslide detection performance
To quantify the effectiveness of the method of the present invention, the landslide detection accuracy and false alarm rate (i.e., the total number of pixels detected as landslides within the area of the geophysical sample other than the landslide divided by the total number of pixels in the sample area) were calculated from the different geophysical pixel samples of the labeled area of FIG. 3, as shown in Table 3.
TABLE 3 landslide detection accuracy and false alarm Rate
Figure BDA0003430966630000151
(3) Analysis discussion of landslide detection result graph
In order to more intuitively analyze the results of different methods of landslide detection, a graph of the results corresponding to table 3 is given in fig. 4.
The range of the landslide region is obtained by comparing Yamaguchi false color images with optical images and visually interpreting, the visual result is an artificially determined result and is used as a judgment standard for judging whether the result obtained by each detection method is correct, the region 1 represents a landslide region, the region 2 represents a field, and the region 3 represents a forest. The method can effectively distinguish landslide from residential areas mainly comprising secondary scattering and forests mainly comprising body scattering, and simultaneously small-area black holes distributed in a landslide area exist, because a large number of large-scale machines (such as excavators, forklifts, earth moving vehicles and the like) exist in the area in disaster relief activities, secondary scattering structures are formed among the mechanical arms, the vehicle body and the ground, so that certain interference is caused on landslide detection, but the determination of a large-area landslide range is not influenced. Compared with a landslide detection method combining change detection and an analytic hierarchy process mentioned in the background technology, the landslide area obtained by the method is more complete and has stronger connectivity, and false alarms of most ground objects in the background are better inhibited, for example, a lower circle of the area 1 represents an area where scattering types before and after landslide do not change, the area can not be detected by the existing detection method to be the landslide area, but the area can be detected by the method disclosed by the invention to be the landslide area. However, the method of the invention still has a small part of false alarms in the field (area 2), which is caused by the growth of crops, and the false alarms brought by the change are eliminated through the coherence map. The forest coverage rate of the area 3 is low, a small amount of bare soil areas are mixed in the middle, and because the areas do not obviously change before and after the landslide, the false alarm at the position can be well inhibited through coherent change detection. Finally, the method is verified to be consistent with a visual result, the existing detection method is likely to detect the terrain features such as fields, bare soil and the like as landslide areas in a wrong way, but the method can eliminate false alarms through coherence detection, and the landslide detection accuracy rate is improved.
Correspondingly to the above method for identifying a complex background downslide based on fusion of polarization information, the present embodiment further provides a device for identifying a complex background downslide based on fusion of polarization information, including:
the change detection module is used for registering images, which have the same polarization combination mode with the pre-landslide single-polarized SAR image, in the pre-landslide single-polarized SAR image and the post-landslide PolSAR image, constructing a double-time phase SAR image pair, and generating a coherent graph through coherent change detection to extract a change area;
the suspected landslide area extraction module is used for sequentially carrying out radiometric calibration, multi-view and terrain correction pretreatment on the post-landslide PolSAR image; extracting a plurality of polarization characteristic parameters from the preprocessed landslide PolSAR image, and fully fusing the polarization characteristic parameters by using TOPSIS (technique for solution analysis with weights) to detect landslides to obtain a region with the same scattering characteristics as the surface of the landslide, namely a suspected landslide region;
and the landslide area determination module is used for fusing the change area extracted from the coherent graph and the suspected landslide area through logical AND operation, and inhibiting false alarm brought by surface features similar to scattering characteristics of the real landslide area in the suspected landslide area through the change area extracted from the coherent graph to obtain a final landslide area.
It should be noted that, in this document, the terms "comprises," "comprising," or any other variation thereof, are intended to cover a non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements does not include only those elements but may include other elements not expressly listed or inherent to such process, method, article, or apparatus.
Finally, it is to be noted that: the above description is only a preferred embodiment of the present invention, and is only used to illustrate the technical solutions of the present invention, and not to limit the protection scope of the present invention. Any modification, equivalent replacement, or improvement made within the spirit and principle of the present invention shall fall within the protection scope of the present invention.

Claims (10)

1. A method for identifying a complex background downslope fused with polarization information is characterized by comprising the following steps:
registering images, which have the same polarization combination mode as the pre-landslide single-polarized SAR image, in the pre-landslide single-polarized SAR image and the post-landslide PolSAR image, constructing a double-time-phase SAR image pair, and generating a coherent image through coherent change detection to extract a change area;
sequentially carrying out radiometric calibration, multi-view and terrain correction preprocessing on the PolSAR image after landslide; extracting multiple polarization characteristic parameters from the preprocessed landslide PolSAR image, and fully fusing the polarization characteristic parameters by using weighted TOPSIS to perform landslide detection to obtain a region with the same scattering characteristics as the landslide surface, namely a suspected landslide region;
and fusing the change area extracted from the coherent graph and the suspected landslide area through logical AND operation, and inhibiting false alarm brought by ground objects with similar scattering characteristics to the real landslide area in the suspected landslide area through the change area extracted from the coherent graph to obtain a final landslide area.
2. The method for identifying the complex background downslope fused with the polarization information as claimed in claim 1, wherein a coherence map is obtained by using a bi-temporal SAR image pair, the magnitude of gray scale values in the coherence map indicates the coherence level, the abrupt nature of the landslide can cause the terrain and landform to change significantly, i.e. the coherence is low, the unchanged area will keep high coherence, and the coherence | γ | is calculated as follows:
Figure FDA0003430966620000011
wherein, E2]Representing a mathematical expectation, S1Is SAR image before landslide, S2The image which is selected from the PolSAR image after the landslide and has the same polarization combination mode as the single-polarized SAR image before the landslide,
Figure FDA0003430966620000012
denotes S2The value range of the coherence | gamma | is [0,1 ]]The larger the | gamma | is, the lower the change degree of a region is, and the smaller the | gamma | is, the more obvious the change of the region is; thresholding and morphological processing of the coherence map using coherence | γ | yields regions of significant variation.
3. The method for identifying complex background downslope fused with polarization information according to claim 1, wherein the method is implemented by YamaGUchi decomposition to obtain surface scattering power psSecondary scattering power pdVolume scattering power pvSum helical scattering power ph(ii) a Obtaining polarization entropy H, average scattering angle alpha and inverse entropy A through H/A/alpha decomposition; calculating the real part Re (rho) of the correlation coefficient of the same polarization component through the correlation coefficienthh-vv)。
4. The method for identifying the downslope of the complex background fused with the polarization information as claimed in claim 3, wherein the power level corresponding to each scattering component, i.e. the surface scattering power P, is extracted from the polarization coherence matrix T by Yamaguchi decompositionsSecondary scattering power PdVolume scattering power PvSum helical scattering power Ph(ii) a The polarization coherence matrix T is represented as a weighted combination of the individual scattered powers:
T＝fsTs+fdTd+fvTv+fhTh
wherein, Ts、Td、Tv、ΤhPolarization coherent matrixes corresponding to surface scattering, secondary scattering, volume scattering and spiral body scattering components are respectively adopted; f. ofs、fd、fv、fhAre respectively Ps、Pd、PvAnd PhThe decomposition coefficient of (a).
5. The method for identifying the complex background downslope fused with the polarization information as claimed in claim 3, wherein the obtaining of polarization entropy H, average scattering angle α and inverse entropy A through H/A/α decomposition comprises: under the condition of single-station reciprocity, the polarization coherent matrix T is a semi-positive definite Hermite matrix of 3 multiplied by 3, so the following eigenvalue decomposition is carried out:
Figure FDA0003430966620000021
wherein the characteristic value lambdaiI is more than or equal to 1 and less than or equal to 3, and satisfies lambda1≥λ2≥λ3>0，λiCorresponding feature vector uiMutually orthogonal, H represents conjugate transpose; the polarization entropy H, the mean scattering angle alpha and the inverse entropy A are determined by the eigenvalues lambdaiAnd a feature vector uiAnd calculating to obtain:
Figure FDA0003430966620000031
in the formula (I), the compound is shown in the specification,
Figure FDA0003430966620000032
for the probability of occurrence of each scattering mechanism,
Figure FDA0003430966620000033
is a feature vector uiThe corresponding scattering angle.
6. The method for identifying the downslope of the complex background fused with the polarization information as claimed in claim 3, wherein the correlation coefficient p of the same polarization componenthh-vvCalculating Re (rho) by a polarization coherence matrix T for a correlation coefficient between two homopolarization scattering vectors of the target, wherein the larger the value is, the more single the scattering type of the target is shown, andhh-vv) The formula of (1) is as follows:
Figure FDA0003430966620000034
wherein Re (. cndot.) represents the real part, T11、T12And T22Representing the corresponding elements in the polarization coherence matrix T.
7. The method for identifying the landslide of complex background with fused polarization information as claimed in claim 3, wherein the landslide detection using weighted TOPSIS fully fused polarization feature parameters comprises the following steps:
firstly, carrying out normalization and homotrending processing on polarization characteristic parameters;
taking the eight polarization characteristic parameters as evaluation indexes, and taking each pixel point of the PolSAR image as an evaluation object to obtain an evaluation matrix;
determining that a positive ideal sample of the evaluation matrix is 1 and a negative ideal sample of the evaluation matrix is 0;
calculating the distance between each pixel point and the positive and negative ideal samples;
calculating the relative closeness of each pixel point to a positive ideal sample;
and determining a proper threshold value to perform threshold segmentation on the relative closeness of each pixel point of the PolSAR image, determining the pixel points with the relative closeness larger than the threshold value as suspected landslide pixel points, and performing morphological processing to obtain a suspected landslide area.
8. The method for identifying the downslope of the complex background fused with the polarization information as claimed in claim 7, wherein the normalizing and homotrending process for the polarization characteristic parameters comprises:
for the scattered power psSecondary scattering power pdVolume scattering power pvSum helical scattering power phAnd (3) carrying out normalization treatment:
Figure FDA0003430966620000041
wherein, the subscript x belongs to { s, d, v, h }, and s, d, v and h respectively represent surface scattering, secondary scattering, volume scattering and spirochete scattering; p after normalization processingd,pv,phCarrying out forward homotrend processing:
px＝1-px
and (3) carrying out interval type normalization processing on the polarization entropy H and the average scattering angle alpha:
Figure FDA0003430966620000042
wherein, for the H, the first and second groups are,
Figure FDA0003430966620000043
in the case of a, for a,
Figure FDA0003430966620000044
real part Re (rho) of the correlation coefficient for the same polarization componenthh-vv) Re (ρ)hh-vv) The negative part of (a) is assigned 0.
9. The method for identifying the downslope of the complex background fused with the polarization information as claimed in claim 7, wherein the AHP method is used to determine the weight of the polarization characteristic parameter, and p is the feature of the landslide after disastersThe surface scattering characteristic of the landslide can be reflected most, H and alpha can better distinguish the landslide from other landforms, namely Re (rho)hh-vv) Surface scattering properties, p, commonly used to reflect topographyd、pv、phAnd A has little effect on landslide identification; thus p will besH and alpha are the most important evaluation indexes, and Re (rho)hh-vv) As a generally important evaluation index, pd、pv、phAnd A as a relatively unimportant evaluation index; and then determining importance comparison values among different evaluation indexes according to the pair comparison matrix importance scale table to construct a judgment matrix.
10. A kind of complicated background that fuses the polarization information slips the slope recognition device, characterized by comprising:
the change detection module is used for registering images, which have the same polarization combination mode with the pre-landslide single-polarized SAR image, in the pre-landslide single-polarized SAR image and the post-landslide PolSAR image, constructing a double-time phase SAR image pair, and generating a coherent graph through coherent change detection to extract a change area;
the suspected landslide area extraction module is used for sequentially carrying out radiometric calibration, multi-view and terrain correction pretreatment on the post-landslide PolSAR image; extracting multiple polarization characteristic parameters from the preprocessed landslide PolSAR image, and fully fusing the polarization characteristic parameters by using weighted TOPSIS to perform landslide detection to obtain a region with the same scattering characteristics as the landslide surface, namely a suspected landslide region;
and the landslide area determination module is used for fusing the change area extracted from the coherent graph and the suspected landslide area through logical AND operation, and inhibiting false alarm brought by surface features similar to scattering characteristics of the real landslide area in the suspected landslide area through the change area extracted from the coherent graph to obtain a final landslide area.
CN202111598126.XA 2021-12-24 2021-12-24 Complex background downslope identification method and device fusing polarization information Pending CN114494282A (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202111598126.XA CN114494282A (en) 2021-12-24 2021-12-24 Complex background downslope identification method and device fusing polarization information

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202111598126.XA CN114494282A (en) 2021-12-24 2021-12-24 Complex background downslope identification method and device fusing polarization information

Publications (1)

Publication Number Publication Date
CN114494282A true CN114494282A (en) 2022-05-13

Family

ID=81496722

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202111598126.XA Pending CN114494282A (en) 2021-12-24 2021-12-24 Complex background downslope identification method and device fusing polarization information

Country Status (1)

Country Link
CN (1) CN114494282A (en)

Cited By (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN115236655A (en) * 2022-09-01 2022-10-25 成都理工大学 Landslide identification method, system, equipment and medium based on fully-polarized SAR
CN116930880A (en) * 2023-07-21 2023-10-24 哈尔滨工业大学 Dynamic evaluation method for deception jamming threat

Cited By (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN115236655A (en) * 2022-09-01 2022-10-25 成都理工大学 Landslide identification method, system, equipment and medium based on fully-polarized SAR
CN115236655B (en) * 2022-09-01 2022-12-20 成都理工大学 Landslide identification method, system, equipment and medium based on fully-polarized SAR
US11747498B1 (en) 2022-09-01 2023-09-05 Chengdu University Of Technology Method, system, device and medium for landslide identification based on full polarimetric SAR
CN116930880A (en) * 2023-07-21 2023-10-24 哈尔滨工业大学 Dynamic evaluation method for deception jamming threat
CN116930880B (en) * 2023-07-21 2024-05-28 哈尔滨工业大学 Dynamic evaluation method for deception jamming threat

Similar Documents

Publication Publication Date Title
Akbari et al. Polarimetric SAR change detection with the complex Hotelling–Lawley trace statistic
Hou et al. Unsupervised change detection in SAR image based on Gauss-log ratio image fusion and compressed projection
Qi et al. A novel algorithm for land use and land cover classification using RADARSAT-2 polarimetric SAR data
Khazai et al. An approach for subpixel anomaly detection in hyperspectral images
Hergoualc'h et al. Characterizing degradation of palm swamp peatlands from space and on the ground: an exploratory study in the Peruvian Amazon
CN103236063B (en) Based on the SAR image oil spilling detection method of multiple dimensioned spectral clustering and decision level fusion
Liu et al. Estimating the soil salinity over partially vegetated surfaces from multispectral remote sensing image using non-negative matrix factorization
Jiang et al. The potential of more accurate InSAR covariance matrix estimation for land cover mapping
CN114494282A (en) Complex background downslope identification method and device fusing polarization information
Qi et al. A three-component method for timely detection of land cover changes using polarimetric SAR images
Su et al. Improving MODIS sea ice detectability using gray level co-occurrence matrix texture analysis method: A case study in the Bohai Sea
CN103971364A (en) Remote sensing image variation detecting method on basis of weighted Gabor wavelet characteristics and two-stage clusters
Myint et al. Urban textural analysis from remote sensor data: Lacunarity measurements based on the differential box counting method
CN115984778A (en) Multi-feature optimization based method for rapidly and dynamically monitoring Sentinel-1 data in flood
Li et al. Object-oriented classification of polarimetric SAR imagery based on statistical region merging and support vector machine
Pirrone et al. A novel framework based on polarimetric change vectors for unsupervised multiclass change detection in dual-pol intensity SAR images
Kseneman et al. Soil-moisture estimation from TerraSAR-X data using neural networks
Feng et al. Detection of urban built-up area change from Sentinel-2 images using multiband temporal texture and one-class random forest
Zhang et al. Superpixelwise likelihood ratio test statistic for PolSAR data and its application to built-up area extraction
Sui et al. Flood detection in PolSAR images based on level set method considering prior geoinformation
Liu et al. Fusing landsat-8, sentinel-1, and sentinel-2 data for river water mapping using multidimensional weighted fusion method
CN114545410B (en) Crop lodging monitoring method based on synthetic aperture radar dual-polarized data coherence
US20220222497A1 (en) Method for identifying dry salt flat based on sentinel-1 data
Liu et al. An Unsupervised Snow Segmentation Approach Based on Dual-Polarized Scattering Mechanism and Deep Neural Network
Wang et al. [Retracted] Remote Sensing Satellite Image‐Based Monitoring of Agricultural Ecosystem

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