Feature point matching landslide toe deformation detection method with terrain constraint added

Abstract

The application discloses a kind of feature point matching landslide rear edge deformation detection methods with terrain constraint, comprising the following steps: first, the image obtained is carried out atmospheric correction, the DN value of image is converted into apparent reflectivity, and the ground reflectivity is obtained by scaling.The feature vector of the feature point used is composed of three parts, the first part is the ground reflectivity of pixel, the second part is the spectral index of pixel, and the third part is the texture feature;Feature point matching;Find point set in target image, so that the feature distance of each pair of matching points is closest;After obtaining the point set, the corresponding matching points are calculated, and the deformation rate of each point on the landslide rear edge is obtained.The advantages of the application are: effectively improve the calculation efficiency and result reliability, effectively improve the utilization rate of remote sensing data and the accuracy of detection results, and realize the automatic monitoring of landslide rear edge deformation movement process.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of slope deformation detection, and particularly relates to a feature point matching landslide rear edge deformation detection method with terrain constraints. BACKGROUND

[0002] The landslide rear edge deformation is one of important indexes for judging the activity of landslide. Because the shape of landslide rear edge is various and the deformation is complex, the detection technology based on multi-period remote sensing images has been difficult to achieve good results. The current method for detecting the landslide rear edge deformation by using optical remote sensing lacks effective use of terrain information and ignores the control effect of the landslide movement law in the detection method, resulting in inaccurate detection. SUMMARY

[0003] The present application provides a feature point matching landslide rear edge deformation detection method with terrain constraints, which comprehensively considers the control effect of terrain on the landslide rear edge in the deformation process of the landslide, greatly improves the speed and accuracy of the feature point selection link, and correspondingly improves the accuracy of the landslide rear edge deformation detection.

[0004] In order to achieve the above application purposes, the technical scheme adopted by the present application is as follows:

[0005] A feature point matching landslide rear edge deformation detection method with terrain constraints comprises the following steps:

[0006] S1: image preprocessing;

[0007] Firstly, the obtained image is subjected to atmospheric correction, the DN (Digital Number) value of the image is converted into apparent reflectivity, and the ground reflectivity is obtained by scaling.

[0008] S2: image feature extraction;

[0009] The feature vector of the feature point comprises three parts, the first part is the ground reflectivity of the pixel, the second part is the spectral index of the pixel, and the third part is the texture feature. The index name and calculation method of the feature vector are shown in Table 1.

[0010] Table 1: Feature and index composition table

[0011]

[0012] Note: Max, Min, Mean, Var represent the maximum value, minimum value, mean value and variance of the data set in the parentheses, respectively.

[0013] S3: feature point matching;

[0014] Let the feature point set labeled by the reference image be X={x1,x2,…,xi ,…,x n}, n is the number of feature points, where the feature vector of the i-th point is x i = [B1, B2, …, B6, RI, MI, VI, Edge]. Let the set of points in the target image that match X be Y = {y1, y2, …, y i ,…,y n}, where the feature vector of the i-th point is y i = [B1, B2, …, B6, RI, MI, VI, Edge]. To achieve the best match of the feature point set, we need to find the set of points Y in the target image such that the feature distance of each pair of matching points is the closest, i.e.

[0015]

[0016]

[0017] The feature distance is calculated using equation (2).

[0018] S4: Deformation feature of the landslide rear edge

[0019] After obtaining the set of points Y that match X, we correspond the matching points one by one, and use the following formula to obtain the deformation rate v i of each point on the landslide rear edge.

[0020]

[0021] where x is the coordinate of the i-th point in X, y is the coordinate of the i-th point in Y, t y and t x are the times when the corresponding images were taken, respectively.

[0022] Further, the following method in S3 is used to find the smallest loss.

[0023] 1) For x i in X, obtain the set of points k in the same position in the target image with a neighborhood radius of r;

[0024] 2) Add the constraint control of the terrain, i.e. filter out the points in k whose elevation is higher than x i , to obtain k′.

[0025] 3) Traverse all points in the set of points k′ and calculate the feature distance with x i , where the closest point is y i ;

[0026] 4) Perform the above three steps for all points in X to obtain the set of points Y = {y1, y2, …, yi ,…,y n}。

[0027] Compared with the prior art, the present application has the advantages of:

[0028] (1) The present application makes full use of multiple characteristic indexes of optical images, combines the law of landslide movement, adds terrain constraints, reduces the number of target point sets, and can effectively improve the calculation efficiency and result reliability.

[0029] (2) For landslide rear edge deformation detection, three new spectral feature indexes are designed on the basis of waveband reflectivity characteristics, which can effectively improve the utilization rate of remote sensing data and the accuracy of detection results.

[0030] (3) The present application can be widely applied to large-scale regional and single-point landslide monitoring, and can quickly and accurately quantitatively calculate the deformation position and rate of the landslide rear edge, realizing the automatic monitoring of the deformation movement process of the landslide rear edge. BRIEF DESCRIPTION OF DRAWINGS

[0031] Figure 1 is an effect diagram of landslide rear edge deformation information extraction at two time points of the embodiment of the present application;

[0032] Wherein, a: satellite image before landslide displacement, b: landslide satellite image after landslide displacement, c: landslide rear edge displacement detection result, d: displacement rate of each feature point, and the satellite image is a true color composite of a Sentinel image. DETAILED DESCRIPTION

[0033] In order to make the purpose, technical scheme and advantages of the present application clearer and more apparent, the following will further describe the present application in detail according to the drawings and examples.

[0034] The basic steps of the feature point matching landslide rear edge deformation detection method with terrain constraints are as follows:

[0035] S1: image preprocessing;

[0036] Firstly, the obtained image is subjected to atmospheric correction, the DN (Digital Number) value of the image is converted into apparent reflectivity, and the surface reflectivity is obtained by scaling, which can be completed by using common commercial software.

[0037] S2: image feature extraction;

[0038] The feature vector of the feature point used in the present method is composed of three parts, the first part is the surface reflectivity of the pixel, which represents the reflection characteristics of each waveband of the pixel, the second part is the spectral index of the pixel, which represents the spectral characteristics between wavebands, and the third part is the texture feature, which is used to describe the spatial distribution characteristics of the pixel. The feature indexes and calculation methods involved in the present method are shown in Table 1.

[0039] Table 1. Characteristic and Index Composition Table

[0040]

[0041]

[0042] Note: Max, Min, Mean, and Var represent the maximum, minimum, mean, and variance of the dataset within the parentheses, respectively.

[0043] S3: Feature point matching;

[0044] Let the feature point set of the reference image be X = {x1, x2, ..., x...} i ,…,x n}, where n is the number of feature points, and the feature vector of the i-th point is x. i = [B1, B2, ..., B6, RI, MI, VI, Edge]. Let Y be the set of points in the target image that match X, where Y = {y1, y2, ..., y...} i ,…,y n}, where the eigenvector of the i-th point is y. i = [B1,B2,…,B6,RI,MI,VI,Edge]. Achieving optimal matching of feature point sets requires finding a point set Y in the target image such that the feature distance between each pair of matching points is minimized, i.e.:

[0045]

[0046]

[0047] The feature distance is calculated using formula (2). This invention proposes the following method to find the minimum loss.

[0048] 1) For x in X i We obtain the set of points k with a neighborhood radius of r in the same position of the target image;

[0049] 2) Considering that landslides can only move from high to low elevations in a time series, the method incorporates topographic constraints, specifically filtering out landslides with elevations higher than x in the k-series. i By finding the point k′, the number of points can be significantly reduced, thus reducing the computation time without affecting the final result.

[0050] 3) Traverse and calculate the relationship between all points in the point set k′ and x. i The feature distance, where the closest point is y. i ;

[0051] 4) Perform the above three steps on all points in X to obtain the point set Y = {y1, y2, ..., y}. i ,…,y n}

[0052] S4: Deformation characteristics at the rear edge of the landslide;

[0053] After obtaining the set of Y points that match the set of X points, the deformation rate v of each point on the trailing edge of the landslide can be obtained by matching the matching points one by one using the following formula. i .

[0054]

[0055] in Let X be the coordinates of the i-th point. Let t be the coordinates of the i-th point in Y. y , t x These represent the times when the corresponding images were captured.

[0056] Validation with example data and result analysis demonstrate that the method is feasible and reliable. Figure 1 Example results of extracting deformation information at the trailing edge of a landslide at two time points.

[0057] Those skilled in the art will recognize that the embodiments described herein are intended to help the reader understand the implementation methods of the present invention, and should be understood that the scope of protection of the present invention is not limited to such specific statements and embodiments. Those skilled in the art can make various other specific modifications and combinations based on the technical teachings disclosed in this invention without departing from the spirit of the invention, and these modifications and combinations are still within the scope of protection of the present invention.

Claims

1. A method for detecting landslide toe deformation by adding terrain-constrained feature point matching, characterized in that, The method comprises the following steps: S1: image preprocessing; Firstly, the acquired image is subjected to atmospheric correction, and DN (Digital Number) value of the image is converted into apparent reflectance, and surface reflectance is obtained through calibration; S2: image feature extraction; The feature vector of the feature point adopted is composed of three parts, the first part is surface reflectance of the pixel, the second part is spectral index of the pixel, and the third part is texture feature, and index name and calculation method of the feature vector are shown in Table 1; Table 1: Feature and index composition table Note: Max, Min, Mean, Var in the table represent maximum value, minimum value, mean value and variance of the data set in the bracket, respectively; S3: feature point matching; Let the feature point set of the reference image be X = {x1, x2,..., x i n, where the feature vector of the ith point is xi = [B1, B2,..., B6, RI, MI, VI, Edge]; let the point set matching X in the target image be Y = {y1, y2,..., y n n, where the feature vector of the ith point is yi = [B1, B2,..., B6, RI, MI, VI, Edge]; i i n i ​​​​ Best matching of the feature point set needs to find a point set Y in the target image, so that the feature distance of each pair of matching points is closest, that is: The feature distance is calculated using formula (2); S4: deformation feature of the landslide rear edge After obtaining the Y point set matched with the X point set, the deformation rate v of each point on the rear edge of the landslide is obtained by using the following formula i ; wherein is the coordinate of the i-th point in X, is the coordinate of the i-th point in Y, t y , t x are the times of the corresponding image captures, respectively.

2. The method according to claim 1, wherein the method is characterized by: The following method is used to find the minimum loss in S3; 1) for x in X i , get the point set k in the target image whose neighborhood radius is r in the same position 2) Adding the constraint control effect of terrain, that is, in screening out the points whose elevation is higher than x in k, k' is obtained. i 2) Adding the constraint control effect of terrain, that is, in screening out the points whose elevation is higher than x in k, k' is obtained. 3) Traverse all points in the point set k' to calculate the distance of each point to x i , where the closest point is y i ; 4) Perform the above three steps on all points in X to get the point set Y = {y1, y2, …, y i ,…,y n}

Images