A landslide detection method, device, equipment and storage medium

Abstract

The embodiments of the present invention provide a landslide detection method, device, equipment and storage medium, which relate to the field of geological disaster prediction technology, including: obtaining pre-seismic and post-seismic remote sensing images of a region to be detected from an optical remote sensing image system and performing registration, performing change detection on the post-seismic remote sensing image based on the registered pre-seismic remote sensing image; extracting a predicted deformation area from the post-seismic non-changing region based on the image of the post-seismic non-changing region and a preset deformation rate map of the region to be detected, and merging it with the post-seismic change region to obtain a target region; calculating the characteristic parameters of each object in the target region based on the grayscale texture feature map calculated from the preset radar image of the region to be detected, the post-seismic remote sensing image, the preset digital elevation model, and the preset deformation rate map, and eliminating the region where the non-landslide features in the target region are located based on the parameters to obtain a landslide hazard region. The present invention can improve the accuracy and comprehensiveness of landslide hazard detection.

Description

Technical Field

[0001] The present invention relates to the technical field of geological disaster prediction, and in particular to a landslide detection method, device, equipment and storage medium. Background Art

[0002] Landslide is the most common geological disaster, which brings huge damage to infrastructure such as roads and houses, seriously threatens people’s lives and property safety, and also causes obvious surface deformation.

[0003] Therefore, timely and accurate landslide detection is essential in geological disaster prediction.

[0004] However, although the conventional change detection method can quickly identify areas of surface changes, it can only identify areas where landslides have occurred, and cannot identify secondary landslides that may occur. Although the use of InSAR technology can identify secondary landslides, decoherence often occurs, resulting in missing data in some areas and data gaps, and requires manual intervention, and is not very accurate. Summary of the invention

[0005] The embodiments of the present invention provide a landslide detection method, device and storage medium, which can improve the accuracy and efficiency of landslide hazard detection and ensure the comprehensiveness of geological analysis.

[0006] In a first aspect, an embodiment of the present invention provides a landslide detection method, comprising:

[0007] Acquire pre-earthquake remote sensing images and post-earthquake remote sensing images of the area to be detected from the optical remote sensing imaging system;

[0008] Using the post-earthquake remote sensing image as a reference, registering the pre-earthquake remote sensing image to the post-earthquake remote sensing image to obtain a registered pre-earthquake remote sensing image;

[0009] Performing change detection on the post-earthquake remote sensing image based on the registered pre-earthquake remote sensing image to obtain an image of the post-earthquake changed area and an image of the post-earthquake non-changed area;

[0010] Extracting a predicted deformation area from the post-earthquake non-changing area according to the image of the post-earthquake non-changing area and the preset deformation rate map of the area to be detected; the coordinate system of the preset deformation rate map is consistent with the coordinate system of the optical remote sensing imaging system;

[0011] Calculating a grayscale texture feature map of a preset radar image according to a grayscale co-occurrence matrix of the preset radar image of the area to be detected; the coordinate system of the preset radar image is consistent with the coordinate system of the optical remote sensing image system;

[0012] Merging the post-earthquake change area and the predicted deformation area to obtain a target area;

[0013] Calculating characteristic parameters of each object in the target area according to the grayscale texture feature map, the post-earthquake remote sensing image, the preset digital elevation model of the area to be detected, and the preset deformation rate map;

[0014] According to the characteristic parameters of each object in the target area, the area where non-landslide objects are located in the target area is eliminated to obtain the landslide potential area.

[0015] Optionally, the registering the pre-earthquake remote sensing image to the post-earthquake remote sensing image based on the post-earthquake remote sensing image to obtain the registered pre-earthquake remote sensing image includes:

[0016] orthorectifying the pre-earthquake remote sensing image and the post-earthquake remote sensing image to obtain a pre-earthquake orthorectified remote sensing image and a post-earthquake orthorectified remote sensing image;

[0017] fusing the panchromatic band data and the multispectral data in the pre-earthquake orthophoto remote sensing image and the post-earthquake orthophoto remote sensing image respectively to obtain a fused pre-earthquake remote sensing image and a fused post-earthquake remote sensing image;

[0018] The fused post-earthquake remote sensing image is used as a reference, and the fused pre-earthquake remote sensing image is registered to the fused post-earthquake remote sensing image as a reference to obtain the registered pre-earthquake remote sensing image.

[0019] Optionally, performing change detection on the post-earthquake remote sensing image based on the registered pre-earthquake remote sensing image to obtain an image of a post-earthquake changed area and an image of a post-earthquake non-changed area includes:

[0020] Performing principal component transformation on the registered pre-earthquake remote sensing image and the post-earthquake remote sensing image respectively to obtain a first principal component image of the registered pre-earthquake remote sensing image and a first principal component image of the post-earthquake remote sensing image; the ground object information in the first principal component image of the registered pre-earthquake remote sensing image corresponds one-to-one with the ground object information of the registered pre-earthquake remote sensing image, and the ground object information in the first principal component image of the post-earthquake remote sensing image corresponds one-to-one with the ground object information of the post-earthquake remote sensing image;

[0021] Determine an area where the position offset between a pixel in the first principal component image of the post-earthquake remote sensing image and a corresponding pixel in the first principal component image of the registered pre-earthquake remote sensing image exceeds a preset offset as a post-earthquake change area in the first principal component image of the post-earthquake remote sensing image;

[0022] According to the post-earthquake change area, determining an image of the post-earthquake change area from the post-earthquake remote sensing image;

[0023] The images of other areas outside the post-earthquake changed area in the post-earthquake remote sensing image are determined as the images of the post-earthquake non-changed area.

[0024] Optionally, extracting a predicted deformation area from the post-earthquake non-changed area according to the image of the post-earthquake non-changed area and the preset deformation information map of the area to be detected includes:

[0025] Calculating the deformation rate value of each pixel in the image of the post-earthquake non-changing area according to the preset deformation rate map;

[0026] The region composed of pixels whose deformation rate values ​​in the image of the post-earthquake non-changing region are within a preset deformation threshold range is determined as the predicted deformation region.

[0027] Optionally, the characteristic parameters of each object include: grayscale texture characteristic value, multispectral value, terrain characteristic value, deformation rate value; the characteristic parameters of each object in the target area are calculated according to the grayscale texture characteristic map, the post-earthquake remote sensing image, the preset digital elevation model of the area to be detected and the preset deformation rate map, including:

[0028] Vectorizing the boundary of the target area to obtain a vector boundary map of the target area;

[0029] Using the vector boundary map, respectively extracting the remote sensing image map of the target area, the digital elevation model of the target area, the deformation rate map of the target area, and the grayscale texture feature map of the target area from the post-earthquake remote sensing image, the preset digital elevation model, the preset deformation rate map, and the grayscale texture feature map;

[0030] Calculating a terrain feature value of each object in the target area according to a digital elevation model of the target area;

[0031] The grayscale texture feature value, the multispectral value and the deformation rate value of each object in the target area are calculated according to the grayscale texture feature map of the target area, the remote sensing image map of the target area and the deformation rate map of the target area.

[0032] Optionally, eliminating the area where non-landslide features are located in the target area according to the characteristic parameters of each object in the target area to obtain the landslide potential area includes:

[0033] A plurality of ground object samples in the remote sensing image of the target area are selected in units of objects; the plurality of ground object samples include all ground object types in the remote sensing image of the target area, and one ground object sample includes only one ground object type;

[0034] According to the characteristic parameters of all objects in the plurality of ground object samples, a preset decision tree algorithm is used to obtain a value range of the characteristic parameters of each ground object type;

[0035] According to the value ranges of the characteristic parameters of the objects of each type of ground object, the corresponding relationship between the value ranges of the characteristic parameters of each object in the target area and each type of ground object is obtained, and a preset ground object classification rule for the target area is generated;

[0036] Using the preset ground object classification rules of the target area, classifying the remote sensing image of the target area to obtain the ground object classification result in the target area;

[0037] According to the result of the land feature classification, the area where the non-landslide land features are located in the target area is eliminated to obtain the landslide potential area.

[0038] Optionally, the terrain characteristic value includes: a slope value, a ground elevation value, a mountain shadow value, and a terrain undulation value. The terrain characteristic value of each object in the target area is calculated based on the digital elevation model of the target area, including:

[0039] Calculating a slope map, a mountain shadow map, a surface undulation map, and a ground elevation map of the target area according to the digital elevation model of the target area;

[0040] The slope value, the hill shadow value, the terrain relief value, and the ground elevation value of each object in the target area are calculated respectively according to the slope map, the hill shadow map, the surface relief map, and the ground elevation map.

[0041] In a second aspect, an embodiment of the present invention further provides a landslide detection device, the device comprising:

[0042] An acquisition module is used to acquire pre-earthquake remote sensing images and post-earthquake remote sensing images of the area to be detected from the optical remote sensing image system;

[0043] A registration module, used to register the pre-earthquake remote sensing image to the post-earthquake remote sensing image based on the post-earthquake remote sensing image, so as to obtain a registered pre-earthquake remote sensing image;

[0044] A detection module, configured to perform change detection on the post-earthquake remote sensing image based on the registered pre-earthquake remote sensing image, to obtain an image of a post-earthquake changed area and an image of a post-earthquake non-changed area;

[0045] An extraction module, configured to extract a predicted deformation area from the post-earthquake non-changing area according to the image of the post-earthquake non-changing area and a preset deformation rate map of the area to be detected; the coordinate system of the preset deformation rate map is consistent with the coordinate system of the optical remote sensing imaging system;

[0046] A first calculation module is used to calculate the grayscale texture feature map of the preset radar image according to the grayscale co-occurrence matrix of the preset radar image of the area to be detected; the coordinate system of the preset radar image is consistent with the coordinate system of the optical remote sensing image system;

[0047] A merging module, used for merging the post-earthquake change area and the predicted deformation area to obtain a target area;

[0048] The second calculation module is further used to calculate the characteristic parameters of each object in the target area according to the grayscale texture feature map, the post-earthquake remote sensing image, the preset digital elevation model of the area to be detected and the preset deformation rate map;

[0049] The elimination module is used to eliminate the area where non-landslide objects are located in the target area according to the characteristic parameters of each object in the target area to obtain the landslide potential area.

[0050] In a third aspect, an embodiment of the present invention further provides a landslide detection device, comprising: a processor, a memory and a bus, wherein the memory stores program instructions executable by the processor, and when the landslide detection device is running, the processor communicates with the memory via the bus, and the processor executes the program instructions to perform the steps of the landslide detection method as described in any one of the first aspects above.

[0051] In a fourth aspect, an embodiment of the present invention further provides a computer-readable storage medium, on which a computer program is stored. When the computer program is executed by a processor, the landslide detection method as described in any one of the first aspects above is executed.

[0052] The present invention provides a landslide detection method, device and storage medium. The method comprises the following steps: registering a pre-seismic remote sensing image and a post-seismic remote sensing image of a region to be detected obtained from an optical remote sensing image system, performing transformation detection on the post-seismic remote sensing image according to the registered pre-seismic remote sensing image, obtaining a post-seismic change region and a post-seismic non-change region, and obtaining a predicted deformation region according to the post-seismic non-change region and a preset deformation rate map. The grayscale texture feature map of the preset radar image is calculated according to the grayscale co-occurrence matrix of the preset radar image of the region to be detected, merging the post-seismic change region and the predicted deformation region to obtain a target region, and then calculating the characteristic parameters of each object in the target region using the grayscale texture feature map, the post-seismic remote sensing image, the preset digital elevation model and the preset deformation rate map. Finally, according to the characteristic parameters of each object in the target region, the region where non-landslide objects are located is eliminated in the target region to obtain a landslide potential area. By this method, change detection is performed on the registered remote sensing image, and the deformation rate map is used to obtain the predicted deformation area, so that the final target area not only has the change area obtained by change detection, but also has the area where deformation has occurred, making the target area information for landslide detection analysis more complete and comprehensive; at the same time, the characteristic parameters of each object in the target area are calculated using the grayscale texture feature map, post-earthquake remote sensing image, preset digital elevation model, and preset deformation rate map, and the area of ​​non-landslide features in the target area is eliminated, so that the final landslide hazard area is more accurate, which can not only provide strong support for landslide detection, but also speed up the efficiency of landslide detection, so that geological detection and analysis can be carried out more efficiently and reliably in the face of landslide disasters, so as to facilitate timely protection of landslide hazard areas. BRIEF DESCRIPTION OF THE DRAWINGS

[0053] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings required for use in the embodiments are briefly introduced below. It should be understood that the following drawings only show certain embodiments of the present invention and therefore should not be regarded as limiting the scope. For ordinary technicians in this field, other related drawings can be obtained based on these drawings without creative work.

[0054] Figure 1 A schematic diagram of a flow chart of a landslide detection method provided by the present invention;

[0055] Figure 2 A schematic diagram of a flow chart of a remote sensing image registration method provided by the present invention;

[0056] Figure 3 A schematic diagram of a flow chart of a change detection method provided by the present invention;

[0057] Figure 4 A schematic diagram of a process for extracting a predicted deformation area provided by the present invention;

[0058] Figure 5 A schematic diagram of a process for calculating characteristic parameters of an object provided by the present invention;

[0059] Figure 6 A schematic diagram of a process for obtaining a landslide hazard area provided by the present invention;

[0060] Figure 7 A schematic diagram of a process for calculating terrain characteristic values ​​provided by the present invention;

[0061] Figure 8 A schematic diagram of a landslide detection device provided by the present invention;

[0062] Fig. 9 A schematic diagram of a landslide detection device provided by the present invention.

[0063] Icons: 1000, acquisition module; 2000, registration module; 3000, detection module; 4000, extraction module; 5000, first calculation module; 6000, merging module; 7000, second calculation module; 8000, elimination module; 10, landslide detection equipment; 11, processor; 12, memory; 13, bus. DETAILED DESCRIPTION

[0064] In order to make the purpose, technical solutions and advantages of the embodiments of the present invention clearer, the technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Generally, the components of the embodiments of the present invention described and shown in the drawings here can be arranged and designed in various different configurations.

[0065] Therefore, the following detailed description of the embodiments of the present invention provided in the accompanying drawings is not intended to limit the scope of the invention claimed for protection, but merely represents selected embodiments of the present invention. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0066] It should be noted that similar reference numerals and letters denote similar items in the following drawings, and therefore, once an item is defined in one drawing, further definition and explanation thereof is not required in subsequent drawings.

[0067] Before explaining the present invention in detail, the application scenario of the present invention is first introduced.

[0068] Landslides mostly occur in mountainous areas, often accompanied by other secondary disasters. However, due to the rugged terrain and inconvenient transportation in mountainous areas, it is difficult to rely on manpower to conduct on-site geological analysis in a short period of time, and the conventional landslide detection methods all have certain limitations. For example, the landslide method based on change detection can quickly obtain the change area, but there is a common phenomenon of over-identification, and the possible secondary landslides are not identified; InSAR (Interferometric Synthetic Aperture Radar, synthetic aperture radar interferometry technology) can identify the hidden dangers of secondary landslides after earthquakes, but the decoherence problem is one of its biggest obstacles; although the object-oriented method can improve the accuracy of ground object classification, the efficiency of landslide identification in a large area is low, and it is not sensitive to the hidden dangers of landslides that have not caused disasters. In addition, the spectral and texture characteristics of optical images can no longer meet the requirements of quantitative and rapid extraction of landslide information. These limitations in landslide detection have led to the inability to provide a reliable landslide detection method to ensure the accuracy of the results.

[0069] Based on this, the present invention proposes a landslide detection method, device and storage medium, by registering the pre-seismic remote sensing image and the post-seismic remote sensing image of the area to be detected obtained from the optical remote sensing image system, transforming and detecting the post-seismic remote sensing image according to the registered pre-seismic remote sensing image, obtaining the post-seismic change area and the post-seismic non-change area, and obtaining the predicted deformation area according to the post-seismic non-change area and the preset deformation rate map, calculating the grayscale texture feature map of the preset radar image according to the grayscale co-occurrence matrix of the preset radar image of the area to be detected, finally merging the post-seismic change area and the predicted deformation area to obtain the target area, and then using the grayscale texture feature map, the post-seismic remote sensing image, the preset digital elevation model, and the preset deformation rate map to calculate the characteristic parameters of each object in the target area, and finally eliminating the area where the non-landslide features are located in the target area according to the characteristic parameters of each object in the target area, and obtaining the landslide hazard area. The landslide detection method provided in the following embodiments of the present invention can be executed by a landslide detection device, which can be a desktop computer, a notebook computer, etc., or a portable intelligent terminal, and the present invention does not limit this.

[0070] The following is an explanation through multiple embodiments in conjunction with the accompanying drawings. Figure 1 The figure is a flow chart of a landslide detection method provided by the present invention. Figure 1 As shown, the landslide detection method comprises:

[0071] S110, obtaining a pre-earthquake remote sensing image and a post-earthquake remote sensing image of the area to be detected from an optical remote sensing image system.

[0072] Remote sensing technology has the advantages of fast imaging speed, wide coverage, and low cost, and is an effective means of disaster and environmental monitoring. An optical remote sensing imaging system is generally composed of a remote sensor, a remote sensing platform, and an information transmission device. The remote sensor is the basis of the optical remote sensing imaging system, and can be a synthetic aperture radar or a multispectral scanner. In this embodiment, drone remote sensing technology can be used to form a panoramic optical remote sensing image of the area to be detected by taking aerial photos of the area to be detected.

[0073] Through the information transmission device, the ground can obtain the pre-earthquake remote sensing images and post-earthquake remote sensing images of the area to be detected taken by the remote sensor from the optical remote sensing image system. Both the pre-earthquake remote sensing images and the post-earthquake remote sensing images are optical remote sensing images.

[0074] S120, using the post-earthquake remote sensing image as a reference, registering the pre-earthquake remote sensing image with the post-earthquake remote sensing image to obtain a registered pre-earthquake remote sensing image.

[0075] Since the acquisition time of the post-earthquake remote sensing image is inconsistent with that of the pre-earthquake remote sensing image, in order to eliminate distortion and prevent the increase of the error rate of subsequent detection due to image misalignment, the two remote sensing images need to be pre-processed. In this embodiment, the post-earthquake remote sensing image is used as a reference, and the pre-earthquake remote sensing image is aligned to the post-earthquake remote sensing image, so that the coordinates of the pre-earthquake remote sensing image are consistent with the coordinates of the post-earthquake remote sensing image, and the aligned pre-earthquake remote sensing image is obtained. Optionally, in the alignment process, the feature points in the two remote sensing images to be aligned can be determined first, and the spatial relationship conversion is performed according to the feature points, so that the pre-earthquake remote sensing image can be aligned to the coordinate system of the post-earthquake remote sensing image according to the conversion matrix, and the aligned pre-earthquake remote sensing image is obtained. In a possible implementation method, the alignment algorithm compiled by the open source program platform can be used to automatically align the two images, and the present invention is not limited to this.

[0076] S130, performing change detection on the post-earthquake remote sensing image based on the registered pre-earthquake remote sensing image to obtain an image of the post-earthquake changed area and an image of the post-earthquake non-changed area.

[0077] In order to obtain the difference between the states observed in different events in the area to be detected, it is necessary to perform transformation detection on the post-earthquake remote sensing image based on the registered pre-earthquake remote sensing image, so as to obtain the image of the post-earthquake changed area and the image of the post-earthquake non-changed area. Generally, according to the selected change detection method, such as algebraic operation method, transformation method, classification method, etc., a suitable threshold can be selected to determine the changed and unchanged images. Among them, the image of the post-earthquake changed area is the image area where obvious surface changes have occurred compared with the registered pre-earthquake remote sensing image. This area is the area that has been affected by the original disaster. Through the results of change detection, the affected area can be locked in time.

[0078] S140, extracting a predicted deformation area from the post-earthquake non-changed area according to the image of the post-earthquake non-changed area and a preset deformation rate map of the area to be detected.

[0079] The coordinate system of the preset deformation rate map is consistent with the coordinate system of the optical remote sensing imaging system.

[0080] In order to improve the accuracy of landslide detection, in addition to focusing on the images of the post-earthquake change area where obvious surface changes have occurred in the change detection, attention should also be paid to the post-earthquake non-change area where secondary disasters may occur. In this embodiment, based on the image of the post-earthquake non-change area and the preset deformation rate map of the area to be detected, the predicted deformation area is extracted from the post-earthquake non-change area, wherein the predicted deformation area is the area in the post-earthquake non-change area where secondary disasters are predicted to occur.

[0081] Specifically, the SAR (Synthetic Aperture Radar) image of the area to be detected is processed by techniques such as Stacking (interference pattern superposition technology), PS (Persistent Scatterer) and SBAS (Small Baseline Subsets) to obtain an unprocessed preset deformation rate map of the area to be detected; then, in order to facilitate subsequent processing, the unprocessed preset deformation rate map of the area to be detected needs to be projected so that the coordinate system of the processed preset deformation rate map is consistent with the coordinate system of the optical remote sensing image system. In a possible implementation, the preset deformation rate map can also be resampled so that the resolution of the preset deformation rate map is consistent with the image resolution in the optical remote sensing image system.

[0082] S150, calculating a grayscale texture feature map of a preset radar image according to a grayscale co-occurrence matrix of a preset radar image of the area to be detected.

[0083] Among them, the coordinate system of the preset radar image is consistent with the coordinate system of the optical remote sensing image system.

[0084] In this embodiment, an unprocessed post-seismic SAR image, that is, an unprocessed post-seismic radar image, can be obtained from the SAR image of the area to be detected. The unprocessed post-seismic radar image is corrected and filtered through an image processing platform, such as the ENVI platform, to obtain a processed post-seismic radar image. Then, based on the post-seismic remote sensing image, the post-seismic radar image is aligned to the post-seismic remote sensing image to obtain an aligned post-seismic radar image. Then, the aligned post-seismic radar image is projected and resampled to make it consistent with the coordinate system and resolution of the optical remote sensing image system to obtain a preset radar image.

[0085] According to the preset radar image, a gray-level co-occurrence matrix (GLCM) of the preset radar image may be calculated.

[0086] Specifically, the statistical law of the grayscale of a pixel (i, j) in a certain direction θ and a certain distance s in the image can be used to reflect the grayscale texture characteristics of the image. This feature is generally represented by a matrix, which is the grayscale co-occurrence matrix.

[0087] In this embodiment, θ is taken as: 0°, 45°, 90°, 135°.

[0088] Assume that the gray level of the preset radar image is N, then the gray level co-occurrence matrix P of the preset radar image is an N×N square matrix. Take any point (x, y) on the preset radar image and the point (Δx, Δy) with a distance s in the θ direction from the point, and set the gray value of point (x, y) to (i, j). Move the point (x, y) across the entire image, and you will get N×N (i, j) values ​​of point (x, y). For the entire image, count the frequency of each (i, j) value, and then arrange them into a square matrix to form the gray level co-occurrence matrix of the preset radar image. At the same time, according to the total number of times (i, j) appears, they are normalized to the probability p(i, j) of the occurrence of this gray value (i, j).

[0089] Then, based on the gray level co-occurrence matrix of the preset radar image, eight gray level texture feature maps of the preset radar image are obtained according to formulas (1) to (8), including: uniformity map, angular second moment map, contrast map, entropy map, correlation map, mean value map, variance sum map, and difference map. The specific process is as follows:

[0090] According to formula (1), the uniformity map of the preset radar image is obtained:

[0091]

[0092] Among them, HOM is homogeneity, which is a measure of the uniformity of the image.

[0093] According to formula (2), the angular second-order moment diagram of the preset radar image is obtained:

[0094]

[0095] Among them, ASM is the angle second moment, which indicates the uniformity of image grayscale texture change and reflects the image grayscale distribution and texture coarseness.

[0096] According to formula (3), the contrast map of the preset radar image is obtained:

[0097]

[0098] Among them, CON is contrast (Contract), which reflects the clarity of image texture.

[0099] According to formula (4), the entropy map of the preset radar image is obtained:

[0100]

[0101] Among them, ENT is entropy, which reflects the amount of information contained in the image.

[0102] According to formula (5), the correlation diagram of the preset radar image is obtained:

[0103]

[0104]

[0105]

[0106] Among them, COR is correlation, which is a measure of the linear relationship of image grayscale and reflects the directionality of image texture.

[0107] According to formula (6), the mean value map of the preset radar image is obtained:

[0108]

[0109] Among them, MEA is the mean value (Mean), generally referred to as the mean, which reflects the regularity of the image texture.

[0110] According to formula (7), the variance and graph of the preset radar image are obtained:

[0111]

[0112] Among them, VAR is the variance sum (Variance), generally referred to as variance, which represents the discrete value of the image brightness value.

[0113] According to formula (8), the difference map of the preset radar image is obtained:

[0114]

[0115] Among them, DIS stands for Dissimilarity, which indicates the size of the difference in image grayscale values.

[0116] Based on the above steps, eight grayscale texture feature maps of the preset radar image can be obtained.

[0117] S160: Merge the post-earthquake change area and the predicted deformation area to obtain a target area.

[0118] By merging the obtained predicted deformation area and the post-earthquake change area, a target area including the post-earthquake change area and the predicted deformation area can be obtained. The target area is the area where landslide hazard detection is required in this embodiment.

[0119] S170, calculating characteristic parameters of each object in the target area according to the grayscale texture feature map, the post-earthquake remote sensing image, the preset digital elevation model of the area to be detected, and the preset deformation rate map.

[0120] Before calculation, the target area needs to be segmented to obtain multiple objects in the target area.

[0121] Specifically, a multi-scale segmentation method may be used to segment the target area, thereby obtaining multiple objects in the target area.

[0122] The quality of the segmentation results directly affects the accuracy of image classification. Among them, the setting of the segmentation scale has a great impact on the segmentation results. In a possible implementation, in order to avoid the segmented objects being too small or too large, the optimal segmentation scale can be determined before segmentation. Segmentation is performed according to the optimal segmentation scale so that the multiple objects in the target area can be effectively distinguished from the ground objects.

[0123] In order to make up for the limitation of landslide detection from optical remote sensing images alone, this embodiment makes a comprehensive judgment based on the characteristics of optical remote sensing images, grayscale texture characteristics of SAR images, terrain characteristics of the area to be detected, and deformation rate characteristics of SAR images.

[0124] First, the unprocessed preset DEM (Digital Elevation Model) of the area to be detected can be downloaded from the high-precision terrain grid data, and the unprocessed preset DEM data can be projected and resampled to make it consistent with the coordinate system and resolution of the optical remote sensing imaging system to obtain the processed preset digital elevation model of the area to be detected.

[0125] Then, the characteristic parameters of each object in the target area can be calculated based on the eight grayscale texture feature maps, post-earthquake remote sensing images, preset digital elevation models of the area to be detected, and preset deformation rate maps. Among them, the characteristic parameters of the object can indicate the characteristics of the optical remote sensing image of the object, the grayscale texture characteristics of the SAR image, the terrain characteristics, and the deformation rate characteristics of the SAR image.

[0126] S180, eliminating the area where non-landslide objects are located in the target area according to the characteristic parameters of each object in the target area, to obtain a landslide potential area.

[0127] According to the characteristic parameters of each object in the target area, the area where non-landslide objects are located is eliminated in the target area, that is, the areas where some objects such as roads, artificial buildings, water bodies, etc. will not cause landslides are eliminated. These areas may be detected as surface changes due to road congestion, house collapse, and water level changes, but they are areas where non-landslide objects are located. By eliminating these areas in the target area, the landslide potential area can be obtained, that is, the area where there is a landslide potential. Specifically, in the area of ​​landslide potential, some areas are areas where primary landslides have occurred and there are secondary landslide risks, and some areas are areas where landslides have not occurred but there are landslide risks due to surface changes. By obtaining the landslide potential area according to the characteristic parameters of each object in the target area, the judgment of the landslide potential area is more accurate, and the landslide potential area can be further protected to prevent the further occurrence of disasters.

[0128] This embodiment performs change detection on the registered remote sensing image and uses the deformation rate map to obtain the predicted deformation area, so that the target area finally obtained not only has the change area obtained by change detection, but also has the area where deformation has occurred, so that the target area information for analyzing landslide detection is more complete and comprehensive; at the same time, the characteristic parameters of each object in the target area calculated by using the grayscale texture feature map, the post-earthquake remote sensing image, the preset digital elevation model, and the preset deformation rate map are eliminated from the area of ​​non-landslide features in the target area, so that the landslide hazard area finally obtained is more accurate, which can not only provide strong support for landslide detection, but also speed up the efficiency of landslide detection, so that when facing landslide disasters, geological detection and analysis can be carried out more efficiently and reliably, so as to facilitate timely protection of landslide hazard areas.

[0129] In the above Figure 1 Based on the provided landslide detection method, the present invention also provides a possible implementation of a remote sensing image registration method. Figure 2 The figure is a flow chart of a remote sensing image registration method provided by the present invention. Figure 2 As shown, in the above S120, the pre-earthquake remote sensing image is registered to the post-earthquake remote sensing image based on the post-earthquake remote sensing image to obtain the registered pre-earthquake remote sensing image, including:

[0130] S122, performing orthorectification on the pre-earthquake remote sensing image and the post-earthquake remote sensing image to obtain the pre-earthquake orthorectified remote sensing image and the post-earthquake orthorectified remote sensing image.

[0131] In order to ensure the accuracy of registration, the pre-earthquake remote sensing images and the post-earthquake remote sensing images need to be orthorectified before registration. For optical remote sensing images, each optical remote sensing image contains panchromatic band data and four-band multispectral data. Specifically, the RPC (Rational Polynomial Coefficients) file and RPC model of the optical remote sensing image data can be used to obtain a complete RPC model; then the preset digital elevation model of the area to be detected and the complete RPC model are used to orthorectify the panchromatic band data and four-band multispectral data of the pre-earthquake remote sensing images and the post-earthquake remote sensing images, respectively, to obtain pre-earthquake orthorectified remote sensing images and post-earthquake orthorectified remote sensing images including panchromatic band data and four-band multispectral data. In a possible implementation method, radiation correction can also be performed on the pre-earthquake remote sensing images and the post-earthquake remote sensing images to eliminate the radiation distortion caused by atmospheric refraction and solar altitude angle.

[0132] S124, respectively fusing the panchromatic band data and the multispectral data in the pre-earthquake orthophoto remote sensing image and the post-earthquake orthophoto remote sensing image to obtain a fused pre-earthquake remote sensing image and a fused post-earthquake remote sensing image.

[0133] After orthorectification, in order to enrich the spectral information of the image, the image needs to be fused. Specifically, a preset image fusion algorithm, such as the NNDiffuse Pan Sharpening algorithm, can be used to fuse the panchromatic band data and multispectral data in the pre-earthquake orthophoto remote sensing image and the post-earthquake orthophoto remote sensing image, respectively, to obtain the fused pre-earthquake remote sensing image and the fused post-earthquake remote sensing image.

[0134] S126, using the fused post-earthquake remote sensing image as a reference, registering the fused pre-earthquake remote sensing image to the fused post-earthquake remote sensing image as a reference, to obtain a registered pre-earthquake remote sensing image.

[0135] After the fusion process, the fused post-earthquake remote sensing image is used as a reference, and the fused pre-earthquake remote sensing image is registered to the fused post-earthquake remote sensing image as a reference to obtain a registered pre-earthquake remote sensing image. The specific registration process is as described in S120 and will not be repeated here.

[0136] In this embodiment, by preprocessing the optical remote sensing image before registration, the registration accuracy can be improved and interference from invalid image information can be avoided.

[0137] In the above Figure 1 Based on the provided landslide detection method, the present invention also provides a possible implementation of a change detection method. Figure 3 The following is a flow chart of a change detection method provided by the present invention. Figure 3As shown, in the above S130, based on the registered pre-earthquake remote sensing image, change detection is performed on the post-earthquake remote sensing image to obtain images of the post-earthquake changed area and images of the post-earthquake non-changed area, including:

[0138] S132, performing principal component transformation on the registered pre-earthquake remote sensing image and the registered post-earthquake remote sensing image, respectively, to obtain a first principal component image of the registered pre-earthquake remote sensing image and a first principal component image of the registered post-earthquake remote sensing image.

[0139] The object information in the first principal component image of the registered pre-earthquake remote sensing image corresponds one-to-one to the object information in the registered pre-earthquake remote sensing image, and the object information in the first principal component image of the post-earthquake remote sensing image corresponds one-to-one to the object information in the post-earthquake remote sensing image.

[0140] Specifically, the feature information includes feature location information and feature type information.

[0141] The principal component transformation can represent a remote sensing image containing more spectral bands with a few bands. The information of the image after principal component transformation is almost not lost, but the data size can be significantly reduced. In this embodiment, the pre-earthquake remote sensing image and the post-earthquake remote sensing image after registration are respectively subjected to principal component transformation, and multiple principal component images of the pre-earthquake remote sensing image and multiple principal component images of the post-earthquake remote sensing image after registration are obtained, and the first principal component image of the pre-earthquake remote sensing image and the first principal component image of the post-earthquake remote sensing image after registration are selected. Among them, the first principal component image is the image contained in the first principal component of the remote sensing image after principal component transformation. By observing the images contained in the first principal component of the two images after principal component transformation, all the ground object information contained in the two images can be obtained with less data.

[0142] S134, determining an area where the position offset between the pixels in the first principal component image of the post-earthquake remote sensing image and the corresponding pixels in the first principal component image of the registered pre-earthquake remote sensing image exceeds a preset offset as a post-earthquake change area in the first principal component image of the post-earthquake remote sensing image.

[0143] Among them, each pixel in the first principal component image of the post-earthquake remote sensing image corresponds to each pixel in the first principal component image of the pre-earthquake remote sensing image after registration; by comparing the position information of each pixel in the first principal component image of the post-earthquake remote sensing image with the position information of each corresponding pixel in the first principal component image of the pre-earthquake remote sensing image after registration, the position offset between each pair of pixels in the two principal component images is obtained; the offset is preset, and the area corresponding to the pixels whose position offset exceeds the preset offset is determined as the post-earthquake change area in the first principal component image of the post-earthquake remote sensing image. In other words, the position offset of the pixel position in the post-earthquake change area in the first principal component image of the post-earthquake remote sensing image exceeds the preset offset compared with the corresponding pixel position in the first principal component image of the pre-earthquake remote sensing image after registration.

[0144] S136, determining an image of the post-earthquake change area from the post-earthquake remote sensing image according to the post-earthquake change area.

[0145] Since the ground object information in the first principal component image of the post-earthquake remote sensing image corresponds one to one with the ground object information in the registered pre-earthquake remote sensing image, the image of the post-earthquake change area can be determined from the post-earthquake remote sensing image based on the post-earthquake change area in the first principal component image of the post-earthquake remote sensing image.

[0146] S138, determining the images of other areas outside the post-earthquake changed area in the post-earthquake remote sensing image as images of the post-earthquake non-changed area.

[0147] After the images of the post-earthquake changed areas are determined, the images of other areas outside the post-earthquake changed areas in the post-earthquake remote sensing images are determined as images of the post-earthquake non-changed areas. That is, the images of the post-earthquake non-changed areas do not have obvious surface deformation compared with the post-earthquake remote sensing images.

[0148] In this embodiment, by using principal component transformation to detect changes in remote sensing images, the amount of image data is reduced while presenting the image's ground object information, reducing the workload and simply, quickly and efficiently analyzing the changed and unchanged areas.

[0149] In the above Figure 1 Based on the provided landslide detection method, the present invention also provides a possible implementation method for extracting a predicted deformation area. Figure 4 A schematic diagram of a process for extracting a predicted deformation region provided by the present invention. Figure 4 As shown, in the above S140, based on the image of the post-earthquake non-changing area and the preset deformation information map of the area to be detected, the predicted deformation area is extracted from the post-earthquake non-changing area, including:

[0150] S142, calculating the deformation rate value of each pixel in the image of the post-earthquake non-changing area according to a preset deformation rate map.

[0151] According to the position of each pixel in the processed deformation rate map of the area to be detected, the position of each pixel in the image of the post-earthquake non-changing area can be corresponded, so as to calculate the deformation rate value of each pixel in the image of the post-earthquake non-changing area.

[0152] S144, determining that the area consisting of pixels whose deformation rate values ​​in the image of the post-earthquake non-changing area are within a preset deformation threshold range is the predicted deformation area.

[0153] According to the deformation rate value of each pixel in the image of the post-earthquake non-changing area, a preset deformation threshold range is set, and the area composed of pixels with deformation rate values ​​within the preset deformation threshold range is extracted as the predicted deformation area. The predicted deformation area is the area in the post-earthquake non-changing area where secondary disasters are predicted to occur.

[0154] In this embodiment, the predicted deformation area is obtained by using a preset deformation rate map and an optical remote sensing image, so that the result of landslide detection is more complete, the identification of potential danger areas is increased, and the problem of failure to detect secondary landslides is avoided.

[0155] In the above Figure 1 Based on the provided landslide detection method, the present invention also provides a possible implementation method for calculating characteristic parameters of an object. Figure 5 The present invention provides a flow chart of calculating object characteristic parameters. Figure 5 As shown, the characteristic parameters of each object include: grayscale texture characteristic value, multispectral value, terrain characteristic value, deformation rate value; in the above S150, the characteristic parameters of each object in the target area are calculated according to the grayscale texture characteristic map, the post-earthquake remote sensing image, the preset digital elevation model of the area to be detected and the preset deformation rate map, including:

[0156] S210, vectorizing the boundary of the target area to obtain a vector boundary map of the target area.

[0157] S220, using the vector boundary map, respectively extracting a remote sensing image map of the target area, a digital elevation model of the target area, a deformation rate map of the target area, and a grayscale texture feature map of the target area from the post-earthquake remote sensing image, the preset digital elevation model, the preset deformation rate map, and the grayscale texture feature map.

[0158] Since the boundary of the target area obtained is very fuzzy and fragmented, in order to obtain a clear boundary of the target area, it is necessary to vectorize the boundary of the target area to facilitate the determination of the boundary and obtain a vector boundary map of the target area. Therefore, the vector boundary map can be used to outline the remote sensing image map of the target area from the post-earthquake remote sensing image, thereby extracting the remote sensing image map of the target area; the vector boundary map is used to outline the remote sensing image map of the target area from the preset digital elevation model, thereby extracting the digital elevation model of the target area; the vector boundary map is used to outline the remote sensing image map of the target area from the preset deformation rate map, thereby extracting the deformation rate map of the target area; the vector boundary map is used to outline the grayscale texture feature map of the target area from the grayscale texture feature map, thereby extracting the grayscale texture feature map of the target area; optionally, if S140 obtains eight grayscale texture feature maps of the preset radar image, then, the vector boundary map can be used to extract eight grayscale texture feature maps of the target area from the eight grayscale texture feature maps.

[0159] After obtaining the remote sensing image of the target area, before performing calculations, the multispectral data in the remote sensing image of the target area can be segmented to obtain multiple objects in the target area.

[0160] Specifically, a multi-scale segmentation method may be used to segment the multispectral data in the remote sensing image of the target area, thereby obtaining multiple objects in the target area.

[0161] S230, calculating a terrain feature value of each object in the target area according to the digital elevation model of the target area.

[0162] First, the terrain characteristic values ​​of all pixels that make up each object in the target area are obtained according to the digital elevation model of the target area, and then the average terrain characteristic values ​​of all pixels that make up each object are calculated respectively. The average terrain characteristic values ​​of all pixels in an object is the terrain characteristic value of the object.

[0163] S240, calculating the grayscale texture feature value, multispectral value and deformation rate value of each object in the target area according to the grayscale texture feature map of the target area, the remote sensing image map of the target area and the deformation rate map of the target area.

[0164] First, according to the grayscale texture feature map of the target area, the remote sensing image map of the target area, and the deformation rate map of the target area, the eight grayscale texture feature values, four-band multispectral values, and deformation rate values ​​of all pixels that make up each object in the target area are calculated. The eight grayscale texture feature values ​​include: uniformity value, angular second moment value, contrast value, entropy value, correlation value, mean value map, variance sum value, and difference value.

[0165] Secondly, the average value of each of the eight grayscale texture eigenvalues ​​of all pixels constituting each object, the average value of the four-band multispectral value, and the average value of the deformation rate are calculated respectively.

[0166] Then, the average value of the grayscale texture feature values ​​of all pixels in an object is the grayscale texture feature value of the object. For example, the average value of the uniformity values ​​of all pixels in an object is the uniformity value of the object; the average value of the four-band multispectral values ​​of all pixels in an object is the four-band multispectral value of the object; the average value of the four-band multispectral values ​​of all pixels in an object is the four-band multispectral value of the object.

[0167] Then, the terrain characteristic value, deformation rate value, multispectral value and grayscale texture characteristic value of each object in the target area are finally obtained, that is, the characteristic parameters of each object in the target area are obtained.

[0168] In this embodiment, by calculating the characteristic parameters of each object, the landslides in the target area can be divided based on the characteristics of the optical remote sensing image of the target area, the grayscale texture characteristics of the SAR image, the terrain characteristics, and the deformation rate characteristics of the SAR image, thereby improving the accuracy of landslide detection.

[0169] In the above Figure 5 Based on the provided method for calculating characteristic parameters of an object, the present invention also provides a possible implementation method for obtaining a landslide hazard area. Figure 6 The present invention provides a flow chart of obtaining a landslide hazard area. Figure 6 As shown, in the above S180, the area where the non-landslide features in the target area are located is eliminated according to the characteristic parameters of each object in the target area to obtain the landslide hazard area, including:

[0170] S310, selecting a plurality of ground object samples in the remote sensing image of the target area in units of objects.

[0171] Wherein, the plurality of ground object samples include all ground object types in the remote sensing image of the target area, and one ground object sample includes only one ground object type;

[0172] There are many types of objects in the remote sensing image of the target area. In order to obtain the object classification rules, visual analysis is required to select multiple object samples in the remote sensing image of the target area by object. For example, sample 1 is selected as a water sample, sample 2 is selected as a mountain sample, and so on. The selected multiple sample types need to include all the types of objects in the remote sensing image of the target area. In addition, it is necessary to ensure that only one type of object is included in the selected single sample.

[0173] S320, according to the characteristic parameters of all objects in the plurality of ground object samples, using a preset decision tree algorithm, obtain a value range of the characteristic parameters of each ground object type.

[0174] The basic unit of each land feature sample is an object, and each object has corresponding characteristic parameters. According to the characteristic parameters of all objects in multiple land feature samples, a preset decision tree algorithm, such as the CART (Classification And Regression Tree) algorithm, can be used to obtain the value range of the characteristic parameters that the objects corresponding to each type of land feature should have.

[0175] S330, obtaining a correspondence between the value range of the characteristic parameter of each object in the target area and each type of ground object according to the value range of the characteristic parameter of each ground object type, and generating a preset ground object classification rule for the target area.

[0176] Since multiple object samples include all types of objects in the remote sensing image of the target area, the corresponding relationship between the value range of the characteristic parameters of each object in the target area and each type of object can be obtained according to the value range of the characteristic parameters of each object type. That is, the value range of the characteristic parameters of each object type can be compared with the characteristic parameter value of each object in the target area to obtain the type of object to which each object belongs, and then the preset object classification rules for the target area are generated. According to the preset object classification rules, the type of object to which any object belongs can be corresponded according to the characteristic parameters of the object, and then for the remote sensing image of the target area, the corresponding positions of each object in this area can be obtained.

[0177] S340, using the preset ground object classification rules of the target area, classifying the remote sensing image of the target area to obtain the ground object classification result in the target area.

[0178] Due to the diverse manifestations of the objects in the target area, in order to accurately divide the final landslide hazard area, it is necessary to classify the remote sensing image of the target area using the preset object classification rules to obtain the object classification results in the target area. For example, the remote sensing image of the target area is classified into water bodies, vegetation, bare land, roads, mountains, artificial buildings and other types of objects.

[0179] S350, according to the result of the land object classification, the area where the non-landslide land objects are located in the target area is eliminated to obtain the landslide potential area.

[0180] According to the classification results of the remote sensing images of the target area, some areas where landslides will not occur, such as roads, artificial buildings, water bodies, etc., are eliminated, and finally the landslide potential areas are obtained.

[0181] In this embodiment, by obtaining the characteristic parameters corresponding to each land object, a preset land object classification rule suitable for each object in the target area can be generated, which can better reflect the land object situation in the target area and avoid misjudgment caused by all visual classifications. By classifying the target land objects, it is convenient to eliminate non-landslide land objects, so that the landslide hazard area can be obtained quickly and effectively, the area of ​​landslide analysis can be reduced, and unnecessary analysis can be omitted.

[0182] In the above Figure 5 On the basis of providing a method for calculating characteristic parameters of an object, the present invention also provides a possible implementation method for calculating a terrain characteristic value. Figure 7 A schematic diagram of a process for calculating terrain characteristic values ​​provided by the present invention. Figure 7 As shown, the terrain characteristic values ​​include: slope value, ground elevation value, mountain shadow value, terrain relief value. In the above S320, after selecting multiple ground object samples in the remote sensing image of the target area in units of objects, it includes:

[0183] S410, calculating and obtaining a slope map, a mountain shadow map, a surface undulation map, and a ground elevation map of the target area according to the digital elevation model of the target area.

[0184] Specifically, the slope map of the target area is calculated according to formula (9);

[0185]

[0186] Among them, slope is the slope, f x is the rate of change of elevation in the X direction in the digital elevation model of the target area, f y is the elevation change rate in the Y direction in the digital elevation model of the target area;

[0187] Calculate the hill shadow map of the target area according to formula (10);

[0188] hillshade=255×(cos(zenith rad )×cos(slope rad ))+

[0189] (sin(zenith rad )×sin(slope rad )×cos(azimuth rad -aspect rad )) (10)

[0190] hillshade∈[0,255],zenith rad , azimuth rad,aspect rad ∈[0,2π]

[0191] Among them, hillshade is the hill shadow value, zenith rad The slope is the solar zenith angle in radians of the remote sensing image. rad is the number of arc degrees of slope in the digital elevation model of the target area, azimuth rad aspect is the arc angle of the sun's rays in the remote sensing image. rad is the number of arc degrees of aspect in the digital elevation model of the target area.

[0192] Calculate the surface relief map and ground elevation map of the target area according to formula (11);

[0193] R=H max -H min ,R>0,H max ,H min ∈R (11)

[0194] Where R is the ground relief, H max is the maximum elevation within the fixed analysis window in the digital elevation model of the target area, H min It is the lowest elevation within the corresponding fixed analysis window in the digital elevation model of the target area.

[0195] S420, respectively calculating the slope value, hill shadow value, terrain relief value, and ground elevation value of each object in the target area according to the slope map, hill shadow map, surface relief map, and ground elevation map.

[0196] Firstly, the slope values, ground elevation values, hill shadow values ​​and terrain relief values ​​of all pixels constituting each object in the target area are obtained according to the slope map, hill shadow map, surface relief map and ground elevation map of the target area.

[0197] Secondly, the average slope value, hill shadow value, surface relief value, and ground elevation value of all pixels constituting each object are calculated respectively. Then, the average slope value of all pixels in an object is the slope value of the object, the average hill shadow value of all pixels in an object is the hill shadow value of the object, the average terrain relief value of all pixels in an object is the terrain relief value of the object, and the average ground elevation value of all pixels in an object is the ground elevation value of the object.

[0198] Then, the slope value, ground elevation value, mountain shadow value, and terrain relief value of each object in the target area are finally obtained.

[0199] By using the above-mentioned characteristic parameters, the characteristic parameters of all objects in multiple ground feature samples can be obtained, and then the value range of the characteristic parameters of each ground feature type object can be obtained by using the preset decision tree algorithm.

[0200] In this embodiment, by using the digital elevation model of the target area to calculate the slope value, ground elevation value, mountain shadow value, and terrain undulation value of each object in the target area, the terrain characteristics of the target area are obtained, which facilitates the use of terrain characteristics to assist in the determination of the landslide area.

[0201] The following is a description of a landslide detection device and a landslide detection equipment provided by the present application for implementation. The specific implementation process and technical effects are as mentioned above and will not be repeated below.

[0202] Figure 8 A schematic diagram of a landslide detection device provided by the present invention is shown in FIG. Figure 8 As shown, the landslide detection device comprises:

[0203] An acquisition module 1000 is used to acquire pre-earthquake remote sensing images and post-earthquake remote sensing images of the area to be detected from an optical remote sensing image system;

[0204] The registration module 2000 is used to register the pre-earthquake remote sensing image with the post-earthquake remote sensing image based on the post-earthquake remote sensing image to obtain the registered pre-earthquake remote sensing image;

[0205] The detection module 3000 is used to perform change detection on the post-earthquake remote sensing images based on the registered pre-earthquake remote sensing images, and obtain images of the post-earthquake changed areas and images of the post-earthquake non-changed areas;

[0206] Extraction module 4000, used to extract the predicted deformation area from the post-earthquake non-changing area according to the image of the post-earthquake non-changing area and the preset deformation rate map of the area to be detected; the coordinate system of the preset deformation rate map is consistent with the coordinate system of the optical remote sensing image system;

[0207] The first calculation module 5000 is used to calculate the grayscale texture feature map of the preset radar image according to the grayscale co-occurrence matrix of the preset radar image of the area to be detected; the coordinate system of the preset radar image is consistent with the coordinate system of the optical remote sensing image system;

[0208] A merging module 6000 is used to merge the post-earthquake change area and the predicted deformation area to obtain a target area;

[0209] The second calculation module 7000 is further used to calculate the characteristic parameters of each object in the target area according to the grayscale texture feature map, the post-earthquake remote sensing image, the preset digital elevation model of the area to be detected, and the preset deformation rate map;

[0210] The elimination module 8000 is used to eliminate the area where non-landslide objects are located in the target area according to the characteristic parameters of each object in the target area, so as to obtain the landslide potential area.

[0211] Optionally, the registration module 2000 is further specifically used to perform orthorectification on the pre-earthquake remote sensing images and the post-earthquake remote sensing images to obtain the pre-earthquake orthorectified remote sensing images and the post-earthquake orthorectified remote sensing images; to fuse the panchromatic band data and the multispectral data in the pre-earthquake orthorectified remote sensing images and the post-earthquake orthorectified remote sensing images respectively to obtain the fused pre-earthquake remote sensing images and the fused post-earthquake remote sensing images; and to register the fused pre-earthquake remote sensing images to the fused post-earthquake remote sensing images based on the fused post-earthquake remote sensing images to obtain the registered pre-earthquake remote sensing images.

[0212] Optionally, the detection module 3000 is further specifically used to perform principal component transformation on the registered pre-seismic remote sensing image and the post-seismic remote sensing image, respectively, to obtain the first principal component image of the registered pre-seismic remote sensing image and the first principal component image of the post-seismic remote sensing image; the ground object information in the first principal component image of the registered pre-seismic remote sensing image corresponds one-to-one with the ground object information of the registered pre-seismic remote sensing image, and the ground object information in the first principal component image of the post-seismic remote sensing image corresponds one-to-one with the ground object information of the post-seismic remote sensing image; the area in which the position offset between the pixel in the first principal component image of the post-seismic remote sensing image and the corresponding pixel in the first principal component image of the registered pre-seismic remote sensing image exceeds a preset offset is determined as the post-seismic change area in the first principal component image of the post-seismic remote sensing image; according to the post-seismic change area, the image of the post-seismic change area is determined from the post-seismic remote sensing image; and the images of other areas outside the post-seismic change area in the post-seismic remote sensing image are determined as the images of the post-seismic non-change area.

[0213] Optionally, the extraction module 4000 is further used to calculate the deformation rate value of each pixel in the image of the post-earthquake non-changed area according to a preset deformation rate map; and determine that the area composed of pixels whose deformation rate values ​​in the image of the post-earthquake non-changed area are within a preset deformation threshold range is the predicted deformation area.

[0214] Optionally, the characteristic parameters of each object include: grayscale texture characteristic value, multispectral value, terrain characteristic value, and deformation rate value.

[0215] Optionally, the extraction module 4000 is also specifically used to vectorize the boundary of the target area to obtain a vector boundary map of the target area; using the vector boundary map, the remote sensing image of the target area, the digital elevation model of the target area, the deformation rate map of the target area, and the grayscale texture feature map of the target area are respectively extracted from the post-earthquake remote sensing image, the preset digital elevation model, the preset deformation rate map, and the grayscale texture feature map.

[0216] Optionally, the second calculation module 7000 is further used to calculate the terrain characteristic value of each object in the target area based on the digital elevation model of the target area; calculate the grayscale texture characteristic value, multispectral value and deformation rate value of each object in the target area based on the grayscale texture feature map of the target area, the remote sensing image map of the target area, and the deformation rate map of the target area.

[0217] Optionally, the elimination module 8000 is also specifically used to select multiple ground object samples in the remote sensing image of the target area in units of objects; the multiple ground object samples include all ground object types in the remote sensing image of the target area, and one ground object sample includes only one ground object type; based on the characteristic parameters of all objects in the multiple ground object samples, a preset decision tree algorithm is used to obtain the value range of the characteristic parameters of each ground object type; based on the value range of the characteristic parameters of each ground object type, the correspondence between the value range of the characteristic parameters of each object in the target area and each ground object type is obtained, and a preset ground object classification rule for the target area is generated; using the preset ground object classification rule of the target area, the remote sensing image of the target area is classified and processed to obtain a ground object classification result in the target area; based on the ground object classification result, the area where non-landslide ground objects are located in the target area is eliminated to obtain a landslide potential area.

[0218] Optionally, the terrain characteristic values ​​include: slope value, ground elevation value, mountain shadow value, and terrain relief value.

[0219] Optionally, the second calculation module 7000 is further specifically used to calculate the slope map, hill shadow map, surface undulation map, and ground elevation map of the target area based on the digital elevation model of the target area; and calculate the slope value, hill shadow value, terrain undulation value, and ground elevation value of each object in the target area based on the slope map, hill shadow map, surface undulation map, and ground elevation map.

[0220] The above modules may be one or more integrated circuits configured to implement the above methods, such as one or more application specific integrated circuits (ASIC), or one or more digital signal processors (DSP), or one or more field programmable gate arrays (FPGA). For another example, when a module is implemented in the form of a processing element scheduling program code, the processing element may be a general-purpose processor, such as a central processing unit (CPU) or other processors that can call program code. For another example, these modules may be integrated together and implemented in the form of a system-on-a-chip (SOC).

[0221] Fig. 9 A schematic diagram of a landslide detection device provided by the present invention, which may be a computing device or server with computing and processing functions.

[0222] The landslide detection device 10 includes: a processor 11, a storage medium 12 and a bus 13. The storage medium 12 stores program instructions executable by the processor 11. When the landslide detection device 10 is executed, the processor 11 communicates with the storage medium 12 through the bus 13. The processor 11 executes the program instructions to execute the above method embodiment. The specific implementation method and technical effect are similar and will not be repeated here.

[0223] Optionally, the present invention also provides a program product, such as a computer-readable storage medium, comprising a program, which is used to execute the above method embodiment when executed by a processor.

[0224] In the several embodiments provided by the present invention, it should be understood that the disclosed devices and methods can be implemented in other ways. For example, the device embodiments described above are only schematic. For example, the division of the units is only a logical function division. There may be other division methods in actual implementation, such as multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed can be through some interfaces, indirect coupling or communication connection of devices or units, which can be electrical, mechanical or other forms.

[0225] The units described as separate components may or may not be physically separated, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed on multiple network units. Some or all of the units may be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0226] In addition, each functional unit in each embodiment of the present invention may be integrated into one processing unit, or each unit may exist physically separately, or two or more units may be integrated into one unit. The above-mentioned integrated unit may be implemented in the form of hardware or in the form of hardware plus software functional units.

[0227] The above-mentioned integrated unit implemented in the form of a software functional unit can be stored in a computer-readable storage medium. The above-mentioned software functional unit is stored in a storage medium, including a number of instructions for a computer device (which can be a personal computer, a server, or a network device, etc.) or a processor (English: processor) to perform some steps of the method described in each embodiment of the present invention. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (English: Read-Only Memory, abbreviated: ROM), random access memory (English: Random Access Memory, abbreviated: RAM), disk or optical disk and other media that can store program codes.

[0228] The above is only a specific embodiment of the present invention, but the protection scope of the present invention is not limited thereto. Any changes or substitutions that can be easily thought of by a person skilled in the art within the technical scope disclosed by the present invention should be included in the protection scope of the present invention. Therefore, the protection scope of the present invention shall be based on the protection scope of the claims.

Claims

1. A landslide detection method, characterized in that: The method comprises: Acquire pre-earthquake remote sensing images and post-earthquake remote sensing images of the area to be detected from the optical remote sensing imaging system; Using the post-earthquake remote sensing image as a reference, registering the pre-earthquake remote sensing image to the post-earthquake remote sensing image to obtain a registered pre-earthquake remote sensing image; Performing change detection on the post-earthquake remote sensing image based on the registered pre-earthquake remote sensing image to obtain an image of the post-earthquake changed area and an image of the post-earthquake non-changed area; Extracting a predicted deformation area from the post-earthquake non-changing area according to the image of the post-earthquake non-changing area and the preset deformation rate map of the area to be detected; the coordinate system of the preset deformation rate map is consistent with the coordinate system of the optical remote sensing imaging system; Calculating a grayscale texture feature map of a preset radar image according to a grayscale co-occurrence matrix of the preset radar image of the area to be detected; the coordinate system of the preset radar image is consistent with the coordinate system of the optical remote sensing image system; Merging the post-earthquake change area and the predicted deformation area to obtain a target area; Calculating characteristic parameters of each object in the target area according to the grayscale texture feature map, the post-earthquake remote sensing image, the preset digital elevation model of the area to be detected, and the preset deformation rate map; Eliminating the area where non-landslide objects are located in the target area according to the characteristic parameters of each object in the target area to obtain a landslide potential area; The characteristic parameters of each object include: grayscale texture characteristic value, multispectral value, terrain characteristic value, and deformation rate value; the characteristic parameters of each object in the target area are calculated according to the grayscale texture characteristic map, the post-earthquake remote sensing image, the preset digital elevation model of the area to be detected, and the preset deformation rate map, including: Vectorizing the boundary of the target area to obtain a vector boundary map of the target area; The vector boundary map is used to extract the remote sensing image of the target area from the post-earthquake remote sensing image; the digital elevation model of the target area is extracted from the preset digital elevation model; the deformation rate map of the target area is extracted from the preset deformation rate map; and the grayscale texture feature map of the target area is extracted from the grayscale texture feature map; Calculating a terrain feature value of each object in the target area according to a digital elevation model of the target area; Calculate the grayscale texture feature value of each object in the target area according to the grayscale texture feature map of the target area; calculate the multispectral value of each object in the target area according to the remote sensing image map of the target area; calculate the deformation rate value of each object in the target area according to the deformation rate map of the target area; Wherein, the step of eliminating the area where non-landslide objects are located in the target area according to the characteristic parameters of each object in the target area to obtain the landslide hazard area includes: A plurality of ground object samples in the remote sensing image of the target area are selected in units of objects; the plurality of ground object samples include all ground object types in the remote sensing image of the target area, and one ground object sample includes only one ground object type; According to the characteristic parameters of all objects in the plurality of ground object samples, a preset decision tree algorithm is used to obtain a value range of the characteristic parameters of each ground object type; According to the value ranges of the characteristic parameters of the objects of each type of ground object, the corresponding relationship between the value ranges of the characteristic parameters of each object in the target area and each type of ground object is obtained, and a preset ground object classification rule for the target area is generated; Using the preset ground object classification rules of the target area, classifying the remote sensing image of the target area to obtain the ground object classification result in the target area; According to the result of the land feature classification, the area where the non-landslide land features are located in the target area is eliminated to obtain the landslide potential area; The terrain characteristic values ​​include: slope value, ground elevation value, mountain shadow value, terrain undulation value, and the terrain characteristic values ​​of each object in the target area are calculated based on the digital elevation model of the target area, including: Calculating a slope map, a mountain shadow map, a surface undulation map, and a ground elevation map of the target area according to the digital elevation model of the target area; The slope value of each object in the target area is calculated according to the slope map of the target area; the hill shadow value of each object in the target area is calculated according to the hill shadow map of the target area; the terrain undulation value of each object in the target area is calculated according to the surface undulation map of the target area; the ground elevation value of each object in the target area is calculated according to the ground elevation map of the target area.

2. The method according to claim 1, characterized in that The method of registering the pre-earthquake remote sensing image with the post-earthquake remote sensing image based on the post-earthquake remote sensing image to obtain the registered pre-earthquake remote sensing image includes: orthorectifying the pre-earthquake remote sensing image and the post-earthquake remote sensing image to obtain a pre-earthquake orthorectified remote sensing image and a post-earthquake orthorectified remote sensing image; fusing the panchromatic band data and the multispectral data in the pre-earthquake orthophoto remote sensing image and the post-earthquake orthophoto remote sensing image respectively to obtain a fused pre-earthquake remote sensing image and a fused post-earthquake remote sensing image; The fused post-earthquake remote sensing image is used as a reference, and the fused pre-earthquake remote sensing image is registered to the fused post-earthquake remote sensing image as a reference to obtain the registered pre-earthquake remote sensing image.

3. The method according to claim 1, characterized in that The step of performing change detection on the post-earthquake remote sensing image based on the registered pre-earthquake remote sensing image to obtain an image of a post-earthquake changed area and an image of a post-earthquake non-changed area includes: Performing principal component transformation on the registered pre-earthquake remote sensing image and the post-earthquake remote sensing image respectively to obtain a first principal component image of the registered pre-earthquake remote sensing image and a first principal component image of the post-earthquake remote sensing image; the ground object information in the first principal component image of the registered pre-earthquake remote sensing image corresponds one-to-one with the ground object information of the registered pre-earthquake remote sensing image, and the ground object information in the first principal component image of the post-earthquake remote sensing image corresponds one-to-one with the ground object information of the post-earthquake remote sensing image; Determine an area where the position offset between a pixel in the first principal component image of the post-earthquake remote sensing image and a corresponding pixel in the first principal component image of the registered pre-earthquake remote sensing image exceeds a preset offset as a post-earthquake change area in the first principal component image of the post-earthquake remote sensing image; According to the post-earthquake change area, determining an image of the post-earthquake change area from the post-earthquake remote sensing image; The images of other areas outside the post-earthquake changed area in the post-earthquake remote sensing image are determined as the images of the post-earthquake non-changed area.

4. The method according to claim 1, characterized in that: The extracting a predicted deformation area from the post-earthquake non-changed area according to the image of the post-earthquake non-changed area and the preset deformation information map of the area to be detected includes: Calculating the deformation rate value of each pixel in the image of the post-earthquake non-changing area according to the preset deformation rate map; The region composed of pixels whose deformation rate values ​​in the image of the post-earthquake non-changing region are within a preset deformation threshold range is determined as the predicted deformation region.

5. A landslide detection device, characterized in that: The device comprises: An acquisition module is used to acquire pre-earthquake remote sensing images and post-earthquake remote sensing images of the area to be detected from the optical remote sensing image system; A registration module, used to register the pre-earthquake remote sensing image to the post-earthquake remote sensing image based on the post-earthquake remote sensing image, so as to obtain a registered pre-earthquake remote sensing image; A detection module, configured to perform change detection on the post-earthquake remote sensing image based on the registered pre-earthquake remote sensing image, to obtain an image of a post-earthquake changed area and an image of a post-earthquake non-changed area; An extraction module, configured to extract a predicted deformation area from the post-earthquake non-changing area according to the image of the post-earthquake non-changing area and a preset deformation rate map of the area to be detected; the coordinate system of the preset deformation rate map is consistent with the coordinate system of the optical remote sensing imaging system; A first calculation module is used to calculate the grayscale texture feature map of the preset radar image according to the grayscale co-occurrence matrix of the preset radar image of the area to be detected; the coordinate system of the preset radar image is consistent with the coordinate system of the optical remote sensing image system; A merging module, used for merging the post-earthquake change area and the predicted deformation area to obtain a target area; The second calculation module is further used to calculate the characteristic parameters of each object in the target area according to the grayscale texture feature map, the post-earthquake remote sensing image, the preset digital elevation model of the area to be detected and the preset deformation rate map; A removal module, used for removing the area where non-landslide objects are located in the target area according to the characteristic parameters of each object in the target area, so as to obtain a landslide potential area; Among them, the characteristic parameters of each object include: grayscale texture characteristic value, multispectral value, terrain characteristic value, deformation rate value; the second calculation module is specifically used to: vectorize the boundary of the target area to obtain the vector boundary map of the target area; use the vector boundary map to extract the remote sensing image of the target area from the post-earthquake remote sensing image; extract the digital elevation model of the target area from the preset digital elevation model; extract the deformation rate map of the target area from the preset deformation rate map; extract the grayscale texture feature map of the target area from the grayscale texture feature map; calculate the terrain characteristic value of each object in the target area according to the digital elevation model of the target area; calculate the grayscale texture characteristic value of each object in the target area according to the grayscale texture feature map of the target area; calculate the multispectral value of each object in the target area according to the remote sensing image map of the target area; calculate the deformation rate value of each object in the target area according to the deformation rate map of the target area; The elimination module is specifically used to: select multiple ground object samples in the remote sensing image of the target area in units of objects; the multiple ground object samples include all ground object types in the remote sensing image of the target area, and one ground object sample includes only one ground object type; according to the characteristic parameters of all objects in the multiple ground object samples, a preset decision tree algorithm is used to obtain the value range of the characteristic parameters of each ground object type object; according to the value range of the characteristic parameters of each ground object type object, a corresponding relationship between the value range of the characteristic parameters of each object in the target area and each ground object type is obtained, and a preset ground object classification rule for the target area is generated; the preset ground object classification rule of the target area is used to classify the remote sensing image map of the target area to obtain a ground object classification result in the target area; according to the ground object classification result, the area where non-landslide ground objects are located in the target area is eliminated to obtain the landslide hazard area; Among them, the terrain feature values ​​include: slope value, ground elevation value, hill shadow value, terrain undulation value, and the second calculation module is specifically used to: calculate the slope map, hill shadow map, surface undulation map, and ground elevation map of the target area according to the digital elevation model of the target area; calculate the slope value of each object in the target area according to the slope map of the target area; calculate the hill shadow value of each object in the target area according to the hill shadow map of the target area; calculate the terrain undulation value of each object in the target area according to the surface undulation map of the target area; calculate the ground elevation value of each object in the target area according to the ground elevation map of the target area.

6. A landslide detection device, characterized in that: include: A processor, a memory and a bus, wherein the memory stores program instructions executable by the processor, and when the landslide detection device is running, the processor communicates with the memory via the bus, and the processor executes the program instructions to perform the steps of the landslide detection method according to any one of claims 1 to 4.

7. A computer-readable storage medium, characterized in that: The storage medium stores a computer program, and when the computer program is executed by the processor, the landslide detection method according to any one of claims 1 to 4 is executed.

Images