Landslide area simulation generation method and device for post-earthquake secondary disasters

Abstract

This disclosure presents a method and apparatus for simulating landslide area generation in post-earthquake secondary disasters. One specific implementation of the method includes: acquiring panchromatic and multispectral image sets, and obtaining original images and elevation datasets; fusing the panchromatic image with each multispectral image in the multispectral image set to generate a fused image; identifying landslide areas on the fused image to generate an identification result; in response to the identification result indicating that the post-earthquake risk area includes landslide areas, performing landslide area edge detection based on the original image, elevation dataset, and fused image to generate a first edge coordinate sequence and a second edge coordinate sequence; and performing landslide area simulation processing based on the original image, fused image, first edge coordinate sequence, and second edge coordinate sequence to generate the landslide surface area and landslide coverage area. This implementation can improve the accuracy of the generated landslide surface area and landslide coverage area.

Description

Technical Field

[0001] The embodiments disclosed herein relate to the field of computer technology, and more particularly to the field of landslide area image recognition technology, specifically to a method and apparatus for simulating and generating landslide area for post-earthquake secondary disasters. Background Technology

[0002] Landslide area is one of the quantitative indicators in landslide hazard assessment, and it is widely used in landslide scale classification, volume estimation, and hazard chain risk assessment. Currently, in estimating the area of ​​landslides after earthquakes, the two-dimensional projected area method (i.e., the planar area measured from orthophotos or maps) is often used to calculate the corresponding landslide area. However, because the two-dimensional projected area method ignores the slope angle between the actual landslide surface and the horizontal plane, the two-dimensional projected area will be systematically smaller than the actual surface area of ​​the landslide on the slope. This is especially true for steep terrain areas (e.g., mountains, canyons), where the steeper the slope, the greater the underestimation of the landslide area, resulting in low accuracy in generating the landslide area. Summary of the Invention

[0003] The summary portion of this disclosure is intended to provide a brief overview of the concepts, which will be described in detail in the detailed description portion. This summary portion is not intended to identify key or essential features of the claimed technical solutions, nor is it intended to limit the scope of the claimed technical solutions.

[0004] Some embodiments of this disclosure propose a method and apparatus for simulating landslide area generation for post-earthquake secondary disasters, in order to solve the technical problems mentioned in the background section above.

[0005] In a first aspect, some embodiments of this disclosure provide a method for simulating landslide area generation for post-earthquake secondary disasters. This method includes: acquiring panchromatic and multispectral image sets, and obtaining original images and elevation datasets. The panchromatic and multispectral images are high-resolution images acquired by a multi-view camera mounted on a UAV at a preset angle targeting the post-earthquake risk area. The original images are color images containing the aforementioned post-earthquake risk area captured by an optical remote sensing satellite before the earthquake time point. The elevation dataset represents the height characteristics of the aforementioned post-earthquake risk area. The method involves fusing the panchromatic images with the various multispectral images in the multispectral image set. The image is processed to generate a fused image; landslide areas are identified in the fused image to generate identification results; in response to the identification results indicating that the post-earthquake risk area includes landslide areas, landslide area edge detection is performed based on the original image, the elevation dataset, and the fused image to generate a first edge coordinate sequence and a second edge coordinate sequence, wherein the first edge coordinate sequence and the second edge coordinate sequence represent the landslide perimeter of the landslide area; landslide area simulation processing is performed based on the original image, the fused image, the first edge coordinate sequence, and the second edge coordinate sequence to generate the landslide surface area and the landslide coverage area.

[0006] Secondly, some embodiments of this disclosure provide a landslide area simulation generation device for post-earthquake secondary disasters. The device includes: an acquisition and processing unit configured to acquire panchromatic images and a multispectral image set, and to acquire original images and an elevation dataset. The panchromatic images and multispectral images are high-resolution images acquired by a multi-view camera mounted on a UAV at a preset angle targeting the post-earthquake risk area. The original images are color images containing the aforementioned post-earthquake risk area taken by an optical remote sensing satellite before the earthquake time point. The elevation dataset represents the height characteristics of the aforementioned post-earthquake risk area. A fusion processing unit is configured to fuse the panchromatic images with each multispectral image in the multispectral image set to generate a fused image. The system includes: a landslide area identification unit configured to identify landslide areas in the fused image to generate identification results; a landslide area edge detection unit configured to, in response to the identification results indicating that the post-earthquake risk area contains landslide areas, perform landslide area edge detection based on the original image, the elevation dataset, and the fused image to generate a first edge coordinate sequence and a second edge coordinate sequence, wherein the first edge coordinate sequence and the second edge coordinate sequence represent the landslide perimeter of the landslide area; and a landslide area simulation processing unit configured to perform landslide area simulation processing based on the original image, the fused image, the first edge coordinate sequence, and the second edge coordinate sequence to generate the landslide surface area and the landslide coverage area.

[0007] Thirdly, some embodiments of this disclosure provide an electronic device, including: one or more processors; and a storage device having one or more programs stored thereon, wherein when the one or more programs are executed by the one or more processors, the one or more processors implement the method described in any implementation of the first aspect above.

[0008] Fourthly, some embodiments of this disclosure provide a computer-readable medium having a computer program stored thereon, wherein the program, when executed by a processor, implements the method described in any of the implementations of the first aspect above.

[0009] The above-described embodiments of this disclosure have the following beneficial effects: the landslide area simulation generation method for post-earthquake secondary disasters according to some embodiments of this disclosure can improve the accuracy of the generated landslide area. Specifically, the landslide area simulation generation method for post-earthquake secondary disasters according to some embodiments of this disclosure firstly acquires panchromatic and multispectral image sets, and obtains original images and elevation datasets. The panchromatic and multispectral images are high-resolution images acquired by a multi-view camera mounted on a UAV at a preset angle targeting the post-earthquake risk area. The original images are color images containing the aforementioned post-earthquake risk area, captured by an optical remote sensing satellite before the earthquake time point. The elevation dataset represents the height characteristics of the aforementioned post-earthquake risk area. Here, the panchromatic image possesses clear geometric texture features, facilitating more accurate extraction of the edge features of the landslide area. Compared to commonly used (red, green, and blue three-channel) color images, multispectral images can provide richer spectral information, allowing for better identification of the landslide area by combining the differences between the landslide area and the surrounding terrain. Furthermore, by introducing the original imagery and elevation dataset, the real topographic environment of the landslide area can be incorporated, thereby improving the accuracy of landslide area localization. Then, the panchromatic imagery is fused with the various multispectral images from the multispectral image set to generate a fused image. This fusion process allows the fused image to possess better texture detail and color features, thus improving the accuracy of landslide area localization. Next, landslide area identification is performed on the fused imagery to generate identification results. This landslide area identification can be used to further determine whether landslide hazards exist within the post-earthquake risk area. Subsequently, in response to the identification results indicating that the post-earthquake risk area contains landslide areas, landslide area edge detection is performed based on the original imagery, the elevation dataset, and the fused imagery to generate a first edge coordinate sequence and a second edge coordinate sequence, where the first and second edge coordinate sequences represent the landslide perimeter of the landslide area. In practice, the accuracy of landslide area identification depends on the precision of landslide boundary positioning. However, because the boundary line of the landslide toe (landslide tongue) is difficult to define, directly identifying the overall boundary of the landslide area can easily lead to significant errors. In contrast, the other locations within the landslide perimeter, except for the toe, have clearly defined surface features due to mountain displacement, making them easier to identify. Therefore, the boundary line of the landslide toe is decoupled from other boundaries, generating a first edge coordinate sequence and a second edge coordinate sequence. This reduces the impact of identifying the toe boundary line on the landslide perimeter identification, thereby improving the positioning accuracy of the landslide boundary. Finally, landslide area simulation processing is performed based on the original image, the fused image, the first edge coordinate sequence, and the second edge coordinate sequence to generate the landslide surface area and landslide coverage area.Here, by combining the real terrain environment (original image) and the landslide area boundary (first edge coordinate sequence and second edge coordinate sequence), a more accurate landslide surface area and landslide coverage area can be generated. Attached Figure Description

[0010] The above and other features, advantages, and aspects of the embodiments of this disclosure will become more apparent from the accompanying drawings and the following detailed description. Throughout the drawings, the same or similar reference numerals denote the same or similar elements. It should be understood that the drawings are schematic, and elements are not necessarily drawn to scale.

[0011] Figure 1 This is a flowchart of some embodiments of the landslide area simulation generation method for post-earthquake secondary disasters disclosed herein;

[0012] Figure 2 These are schematic diagrams of landslide areas in some embodiments;

[0013] Figure 3 These are schematic diagrams of landslide areas from other embodiments;

[0014] Figure 4 This is a structural schematic diagram of some embodiments of the landslide area simulation generation device for post-earthquake secondary disasters according to the present disclosure;

[0015] Figure 5 This is a schematic diagram of the structure of an electronic device suitable for implementing some embodiments of the present disclosure. Detailed Implementation

[0016] Embodiments of this disclosure will now be described in more detail with reference to the accompanying drawings. While some embodiments of this disclosure are shown in the drawings, it should be understood that this disclosure can be implemented in various forms and should not be construed as limited to the embodiments set forth herein. Rather, these embodiments are provided to provide a more thorough and complete understanding of this disclosure. It should be understood that the accompanying drawings and embodiments of this disclosure are for illustrative purposes only and are not intended to limit the scope of protection of this disclosure.

[0017] It should also be noted that, for ease of description, only the parts relevant to the invention are shown in the accompanying drawings. Unless otherwise specified, the embodiments and features described in this disclosure can be combined with each other.

[0018] It should be noted that the concepts of "first" and "second" mentioned in this disclosure are used only to distinguish different devices, modules or units, and are not used to limit the order of functions performed by these devices, modules or units or their interdependencies.

[0019] It should be noted that the terms "a" and "a plurality of" used in this disclosure are illustrative rather than restrictive, and those skilled in the art should understand that, unless otherwise expressly indicated in the context, they should be understood as "one or more".

[0020] The names of messages or information exchanged between multiple devices in the embodiments of this disclosure are for illustrative purposes only and are not intended to limit the scope of such messages or information.

[0021] This disclosure will now be described in detail with reference to the accompanying drawings and embodiments.

[0022] Figure 1 A flowchart 100 is shown, illustrating some embodiments of a landslide area simulation generation method for post-earthquake secondary disasters according to this disclosure. The landslide area simulation generation method for post-earthquake secondary disasters includes the following steps:

[0023] Step 101: Acquire panchromatic and multispectral image sets, and obtain raw image and elevation datasets.

[0024] In some embodiments, the execution entity (e.g., a computing device) of the landslide area simulation generation method for post-earthquake secondary disasters can wirelessly acquire panchromatic and multispectral image sets, as well as obtain raw images and elevation datasets. The panchromatic and multispectral images are high-resolution images acquired at preset angles by multi-view cameras mounted on a drone, targeting the post-earthquake risk area. The raw images are color images containing the aforementioned post-earthquake risk area, captured by optical remote sensing satellites before the earthquake time point. The elevation dataset characterizes the height features of the aforementioned post-earthquake risk area. The panchromatic image is a single-channel grayscale image containing the post-earthquake risk area. The multispectral image in the multispectral image set is a single-band image with discrete spectra. Furthermore, the post-earthquake risk area can be a pre-selected suspected landslide area. The earthquake time point is the starting time of the earthquake.

[0025] In practice, generating landslide areas through image recognition relies heavily on image clarity and texture. This is particularly problematic in post-earthquake secondary disaster areas where road and infrastructure are often damaged. Using remote sensing satellites to photograph landslide areas can lead to unstable image clarity due to weather conditions, compromising the accuracy of the generated landslide areas. Therefore, a drone equipped with a multi-view camera is used to acquire panchromatic and multispectral images of the post-earthquake risk area at a preset angle. This preset angle can be a pre-defined angle directly facing the post-earthquake risk area (e.g., 45 degrees downwards from the risk area) to avoid the influence of vegetation on landslide boundary identification due to angle distortion. Furthermore, since landslides often involve the rolling of debris (e.g., rocks), resulting in a large affected area, the panchromatic and multispectral images can cover a wider area than the post-earthquake risk area.

[0026] Secondly, landslides are essentially geological processes occurring on slopes, and their actual distribution area is inevitably controlled by the slope's inclination. However, the two-dimensional projection area method ignores the slope angle, causing the calculated landslide area to deviate from the actual terrain environment and reducing the accuracy of the generated landslide area. Therefore, by acquiring original images and elevation datasets, the real terrain environment of the landslide area is incorporated to improve the accuracy of the landslide area calculation. Here, since it is difficult to determine the specific landslide area before the earthquake, color images containing the aforementioned post-earthquake risk areas are selected as the original images to incorporate the real terrain environment of the landslide area. Furthermore, by limiting the original images to a preset time period before the earthquake (e.g., within 10 days), the significant differences in the surface environment (e.g., vegetation) caused by large time differences in the original images can be avoided, preventing the introduction of more identification errors.

[0027] It should be noted that the aforementioned wireless connection methods may include, but are not limited to, 3G / 4G / 5G connections, WiFi connections, Bluetooth connections, WiMAX connections, Zigbee connections, UWB (Ultra Wide Band) connections, and other currently known or future wireless connection methods.

[0028] It should be noted that the aforementioned computing devices can be either hardware or software. When the computing device is hardware, it can be implemented as a distributed cluster consisting of multiple servers or terminal devices, or as a single server or terminal device. When the computing device is software, it can be installed within the hardware devices listed above. It can be implemented as, for example, multiple software programs or software modules used to provide distributed services, or as a single software program or software module. No specific limitations are made here. It should be understood that the number of computing devices can be arbitrary, depending on the implementation requirements.

[0029] In some optional implementations of certain embodiments, the aforementioned execution entity acquires the original imagery and elevation dataset, including the following steps:

[0030] Step S1: Select historical image maps from the historical image map library that contain the aforementioned post-earthquake risk areas and meet the preset filtering criteria, and use them as target image maps. The target image map is a map of the original terrain containing the post-earthquake risk areas before the earthquake. The preset filtering criteria are that the time point corresponding to the historical image map is before the earthquake time point, and the time interval between the two is the shortest. Here, the historical image map library is used to store historical image maps observed at different time points.

[0031] Step S2: Extract the area corresponding to the post-earthquake risk area from the target image map as the original image. The image corresponding to the post-earthquake risk area and the target image map can be taken by the same optical remote sensing satellite. Therefore, the post-earthquake risk area can be magnified in the target image map at a preset magnification ratio (e.g., 1.5 times), using the center point of the post-earthquake risk area as a fixed coordinate, to obtain the area corresponding to the original image. In practice, because the surface structure of landslide areas differs from adjacent areas, a larger area needs to be determined to further confirm whether the post-earthquake risk area is indeed a landslide area, avoiding false detections of the post-earthquake risk area.

[0032] Step S3: Select the elevation data corresponding to the original image from the elevation database to obtain the elevation dataset. The elevation database can be a database of pre-measured height values ​​for each coordinate in the target image map. Therefore, the elevation data corresponding to each coordinate in the original image can be selected from the elevation database to obtain the elevation dataset. The elevation data represents the surface height value at the coordinate location.

[0033] Step 102: The panchromatic image is fused with each multispectral image in the multispectral image group to generate a fused image.

[0034] In some embodiments, the executing entity may fuse the panchromatic image with each multispectral image in the multispectral image group to generate a fused image. The fused image is a color image that incorporates features from both the panchromatic and multispectral images. Compared to commonly used, lower-resolution multispectral images, the fused image improves the spatial resolution of each channel and provides more pronounced texture features. Specifically, each multispectral image in the multispectral image group may correspond to one of the four bands: red, green, blue, and near-infrared.

[0035] Secondly, to better fuse the texture features of the panchromatic image and the color features of the multispectral image, the NNDiffuse (Nearest Neighbor Diffusion) fusion algorithm is used to fuse the panchromatic image and each multispectral image in the multispectral image group to generate a fused image. Specifically, the NNDiffuse fusion algorithm has advantages such as high spectral fidelity, fast computation efficiency, and strong tolerance to registration errors. It can effectively inject high-resolution spatial texture (i.e., the texture features of the panchromatic image) without changing the original multispectral reflectance. This allows the fused image to simultaneously utilize the fine features of the landslide boundary and the NDVI (Normalized Difference Vegetation Index) features of the multispectral image corresponding to the near-infrared band, thus balancing the accuracy and efficiency of area calculation.

[0036] Optionally, the panchromatic image and each multispectral image can be fused using the GS (Gram-Schmidt) orthogonalization fusion algorithm to obtain a fused image.

[0037] Step 103: Identify landslide areas in the fused image to generate identification results.

[0038] In some embodiments, the aforementioned execution entity can perform landslide area identification on the fused image to generate an identification result. First, the fused image is clustered using a mean-shift algorithm to obtain sub-region groups. Here, sub-regions are formed by merging neighboring regions with similar color. Second, for each sub-region, the following steps are performed: The vegetation normalization index of the sub-region can be determined as the post-earthquake normalization index. The vegetation normalization index corresponding to the sub-region in the original image is obtained as the pre-earthquake normalization index. Here, the vegetation normalization index corresponding to the original image can be pre-generated. The sub-regions of the original image can be obtained by projecting the boundary coordinates of the sub-regions generated by region clustering onto the image coordinate system of the original image through coordinate transformation. Then, the difference between the post-earthquake normalization index and the pre-earthquake normalization index is determined as the vegetation index difference. Finally, if there is a vegetation index difference greater than a preset index threshold, an identification result representing a landslide area within the post-earthquake risk area is generated.

[0039] Optionally, landslide area identification can also be performed on the fused image using pre-trained semantic segmentation to generate identification results. The semantic segmentation model can be a standard U-shaped network structure. The semantic segmentation model can include an encoder and a decoder. The encoder includes four downsampling layers, and the decoder includes four upsampling layers, with skip connections between the upsampling and downsampling layers.

[0040] Step 104: In response to the identification result indicating that the post-earthquake risk area includes a landslide area, landslide area edge detection is performed based on the original image, elevation dataset and fused image to generate a first edge coordinate sequence and a second edge coordinate sequence.

[0041] In some embodiments, the execution entity may, in response to the identification result indicating that the post-earthquake risk area includes a landslide area, perform landslide area edge detection based on the original image, the elevation dataset, and the fused image to generate a first edge coordinate sequence and a second edge coordinate sequence. The first edge coordinate sequence and the second edge coordinate sequence represent the landslide perimeter of the landslide area.

[0042] As an example, see Figure 2 The boundary 201 corresponding to the first edge coordinate sequence represents the dividing line between the area where the landslide wall 202 is located and the landslide body 203.

[0043] In practice, the accuracy of landslide area identification depends on the precision of landslide boundary positioning. However, since the boundary line at the toe of a landslide is difficult to define, directly identifying the overall boundary of the landslide area can easily lead to significant errors. Therefore, the boundary line at the toe of the landslide is decoupled from other boundaries within the landslide perimeter, generating a first edge coordinate sequence and a second edge coordinate sequence respectively. This reduces the impact of identifying the boundary line at the toe of the landslide on the identification of the landslide perimeter, thereby improving the positioning accuracy of the landslide boundary.

[0044] Secondly, an active contour model is first used to extract the initial edge coordinate sequence from the fused image. The active contour model is an image segmentation algorithm based on energy minimization. Specifically, the active contour model does not directly rely on pixel-level "edge" judgment, but rather smoothly delineates complete and closed boundaries through iterative optimization, effectively filtering noise interference (e.g., cracks near landslide boundaries). Combining the active contour model with digital terrain features enables efficient and accurate identification of landslide boundaries.

[0045] Then, based on the pre-established coordinate transformation relationship between the optical remote sensing satellite and the UAV, the initial edge coordinates in the initial edge coordinate sequence are transformed from the image coordinate system of the fused image to the original image through coordinate transformation, resulting in the current transformed coordinate sequence. Next, elevation data corresponding to the current transformed coordinates is selected from the elevation dataset as the vertical coordinate value of the current transformed coordinates, resulting in a three-dimensional transformed coordinate sequence. Then, the point with the lowest vertical coordinate value is selected as the first target transformed coordinate, and the point with the highest vertical coordinate value is selected as the second target transformed coordinate. Next, the three-dimensional transformed coordinate sequence is divided into two groups according to the first and second target transformed coordinates, resulting in the first coordinate group and the second coordinate group. Then, each first coordinate in the first coordinate group is fitted to obtain the first coordinate curve equation. Each second coordinate in the second coordinate group is fitted to obtain the second coordinate curve equation. Then, the point with the largest curvature in the first coordinate curve equation is selected as the first curvature abrupt change point. The point with the largest curvature in the second coordinate curve equation is selected as the second curvature abrupt change point. Here, the first curvature abrupt change point and the second curvature abrupt change point characterize the boundary points of the landslide curvature abrupt change (i.e., the boundary positions where the landslide body changes from a tilted sliding state to a gentle sliding state). Therefore, the three-dimensional transformation coordinate sequence can be divided into two groups according to the first curvature abrupt change point and the second curvature abrupt change point, which are respectively used as the first edge coordinate sequence and the second edge coordinate sequence.

[0046] In some optional implementations of certain embodiments, the execution entity performs landslide area edge detection based on the original image, the elevation dataset, and the fused image to generate a first edge coordinate sequence and a second edge coordinate sequence, including:

[0047] Step S1 involves calibrating the original imagery and elevation dataset to obtain calibrated imagery and calibrated elevation datasets. Since the original imagery and elevation datasets originate from different sensors, shooting times, or coordinate systems, and each possesses geometric deformations and errors, calibration is necessary. Secondly, a calibration model based on RPC (Rational Polynomial Coefficients) can be used to calibrate the original imagery and elevation datasets, resulting in calibrated imagery and calibrated elevation datasets. Here, the RPC-based calibration model performs orthorectification on the original imagery and each multispectral image separately, ensuring accurate geometric alignment of coordinates within the images. The elevation dataset serves as a vertical reference during the calibration process, correcting geometric deformations caused by terrain and eliminating local offsets caused by engineering structures or vegetation. Furthermore, the transformation relationship between the original imagery and the calibrated imagery (such as the orthorectification model) can be saved for coordinate transformation.

[0048] Step S2 involves edge detection of the fused image to generate an initial coordinate sequence. First, the fused image can be edge-featured using the Canny operator (a multi-level edge detection operator) + OTSU (Nobuyuki Otsu, image thresholding) algorithm to generate an edge feature coordinate sequence. Here, the OTSU algorithm is used to calculate the high threshold of the Canny operator, and the ratio of the low to high thresholds of the Canny operator can be [1:2, 1:3]. Second, an edge detection map can be generated. Here, the edge detection map has the same scale as the fused image. The edge detection map is a binary image. Pixels corresponding to edge feature coordinates in the edge detection map have a value of 1, while pixels in other areas have a value of 0. Then, a secondary edge detection can be performed on the fused image using a Transformer-based edge detection network (e.g., a pre-trained Swin-Transformer deep learning model) to generate the initial coordinate sequence. Here, the layers corresponding to each spectrum of the fused image (i.e., red band, green band, blue band, and near-infrared band) and the edge detection map can be used as a 5-channel tensor as input to the edge detection network. The edge detection network consists of a multi-scale feature extraction layer, a spatial attention module, a Transformer encoding module, and an output layer. Here, the multi-scale feature extraction layer quickly locates the edge based on the geometric contour features in the edge detection image. The spatial attention module enhances the contour features of the landslide edge and suppresses false edges caused by debris, shadows, and vegetation. The Transformer encoding module, based on a self-attention mechanism, calculates the correlation weights between all pixels in the image to capture long-distance feature dependencies globally, perfectly suited to the edge fracturing problem in earthquake-induced landslide scenarios, for completing the landslide edge and obtaining a landslide edge coordinate sequence and the confidence score corresponding to each coordinate. Finally, landslide edge confidence scores greater than a preset coordinate confidence threshold are used as initial coordinates to obtain an initial coordinate sequence. In practice, since the rear edge of a landslide and the sidewalls on both sides of the landslide often have obvious geographical feature changes (such as color changes), the confidence level of the landslide edge coordinates corresponding to the rear edge of the landslide and the sidewalls on both sides of the landslide is often high. However, the landslide tongue boundary is blurred, so the confidence level of the landslide edge coordinates corresponding to the landslide tongue is often low. Therefore, by using confidence level screening, a portion of the landslide perimeter (initial coordinate sequence) that represents the rear edge of the landslide and the sidewalls on both sides of the landslide can be selected.

[0049] In practice, in post-earthquake landslide scenarios, directly using a model for end-to-end edge detection is constrained by technical issues such as dense debris on the post-earthquake surface, gradually blurred landslide boundaries, uneven image brightness distribution, and scarce disaster samples. This can easily lead to defects such as false edge interference from rocks, missing boundary fractures, and insufficient generalization. However, pre-generating an initial edge feature layer using the Canny multi-level edge detection operator combined with the OTSU adaptive thresholding algorithm for edge extraction can filter out fine noise from debris while preserving the continuous contour of the entire landslide area. This initial edge layer, stitched together with the fused image channels and input into the MAST model, provides the model with explicit geometric contour priors, reducing the difficulty of edge learning under complex surfaces, effectively suppressing false edge interference caused by earthquake disturbances, repairing landslide boundary fractures, and resulting in continuous, high-precision edge coordinates that conform to the geological morphology of the landslide.

[0050] Step S3: Transform the initial coordinates in the initial coordinate sequence from the image coordinate system corresponding to the fused image to the image coordinate system corresponding to the calibrated image to obtain the transformed coordinate sequence. Specifically, based on a pre-established coordinate transformation relationship, the initial coordinates can be transformed from the image coordinate system corresponding to the fused image to the image coordinate system corresponding to the original image. Then, based on the transformation relationship between the original image and the calibrated image, transform the initial coordinates in the image coordinate system corresponding to the original image to the image coordinate system corresponding to the calibrated image to obtain the transformed coordinates.

[0051] Step S4 involves transforming the transformed coordinates in the above-mentioned transformed coordinate sequence from the image coordinate system corresponding to the calibrated image to the planar geographic coordinate system, thus obtaining the planar geographic coordinate sequence. The planar geographic coordinate system is the projected coordinate system. Here, the transformed coordinates can be transformed from the image coordinate system corresponding to the calibrated image to the planar geographic coordinate system based on a pre-established mapping relationship between optical remote sensing satellites and projected coordinate systems, thus obtaining the planar geographic coordinates.

[0052] In practice, while fused imagery combines high resolution and multispectral information, its spatial resolution is still limited by the original panchromatic image. Directly using the pixel coordinates of the original panchromatic image can achieve sub-meter or even centimeter-level edge localization accuracy, avoiding resampling errors during the fusion process. Secondly, the fusion process may introduce minute local geometric distortions; transforming the coordinates to the image coordinate system of the calibrated image can significantly eliminate these distortions, thereby eliminating fusion errors and locking in high-resolution pixels. Then, by generating planar geographic coordinates, the correspondence between the coordinates and the actual geographic environment can be determined, facilitating subsequent measurements.

[0053] Step S5: For each planar geographic coordinate in the above-mentioned planar geographic coordinate sequence, determine the vertical coordinate value of the above-mentioned planar geographic coordinate system based on the above-mentioned calibrated elevation dataset to obtain the first edge coordinate. Specifically, the correspondence between the planar geographic coordinates and the calibrated elevation data can be used to select the elevation data corresponding to the above-mentioned planar geographic coordinate system as the vertical coordinate value of the above-mentioned planar geographic coordinate to obtain the first edge coordinate.

[0054] Step S6: Generate a second edge coordinate sequence based on the obtained first edge coordinate sequence. Specifically, the third-dimensional transformed coordinates in the above-mentioned three-dimensional transformed coordinate sequence whose distance from the first edge coordinates in the above-mentioned first edge coordinate sequence is greater than a preset interval threshold can be used as the second edge coordinates to obtain the second edge coordinate sequence.

[0055] Optionally, the execution entity generates a second edge coordinate sequence based on the obtained first edge coordinate sequence, including:

[0056] Step S1: Based on the first edge coordinate sequence, perform image segmentation on the calibrated image and the fused image respectively to obtain the segmented original image and the segmented fused image. First, determine the first edge coordinate with the lowest vertical coordinate value in the first edge coordinate sequence as the lowest point coordinate. Then, determine the coordinates corresponding to the lowest point coordinate in the calibrated image and the fused image respectively to obtain the first initial coordinate and the second initial coordinate. For the calibrated image, extend the boundary of the transformed coordinate sequence horizontally along the first initial coordinate position towards the adjacent image side boundary to obtain the first segmentation line. Thus, the original image can be segmented according to the first segmentation line; here, the segmented sub-image representing the landslide area in the calibrated image is used as the segmented original image. For the fused image, extend the boundary corresponding to each initial coordinate in the initial coordinate sequence horizontally along the first initial coordinate position towards the adjacent image boundary to obtain the second segmentation line. Thus, the fused image can be segmented according to the second segmentation line; here, the segmented sub-image representing the landslide area in the fused image is used as the segmented fused image. In practice, image segmentation can remove areas outside the landslide area from an image, thereby reducing computational load in subsequent recognition processes and avoiding the negative impact of areas outside the landslide area on landslide area recognition.

[0057] Step S2 involves identifying scattered blocks in the segmented original image to generate a first set of scattered block coordinates. This first set of scattered block coordinates represents the location distribution of scattered blocks on the surface before the earthquake time point. Next, a pre-trained YOLO v8 (You Only Look Once, Object Detection) model is used to identify the segmented original image to generate a first set of scattered block identification information. Each piece of first scattered block identification information includes a scattered block detection box. Finally, the geometric center of the scattered block detection box in the first scattered block identification information is used as the first scattered block coordinates to obtain the first set of scattered block coordinates.

[0058] In practice, landslides in landslide areas are often accompanied by the scattering of rocks and other boulders. These scattering blocks, accelerated to a certain height, possess significant kinetic energy and are highly destructive. Furthermore, due to the extremely complex topography and geology, the boundaries of landslide tongues after a landslide are highly irregular, making it difficult to ensure the accuracy of landslide tongue boundary identification. Therefore, this disclosure transforms the complex problem of landslide tongue boundary delineation into the inclusion of the area covered by highly destructive scattering within the landslide area. This not only ensures that the landslide coverage area includes the radiating area of ​​the scattering but also avoids the technical challenge of identifying landslide tongue boundaries, thereby improving the accuracy of the generated landslide coverage area.

[0059] Step S3: Scattered block identification is performed on the segmented and fused image to generate a second scattered block coordinate set. This second scattered block coordinate set represents the location distribution of scattered blocks on the surface after the earthquake time point. Next, the segmented and fused image is first identified using the YOLO v8 (You Only Look Once, Object Detection) model to generate a second scattered block identification information set. Each piece of second scattered block identification information includes a scattered block detection box. Finally, the geometric center of the scattered block detection box in the second scattered block identification information is used as the second scattered block coordinates to obtain the second scattered block coordinate set.

[0060] Step S4: Transform the coordinate set of the second scattered block into the image coordinate system corresponding to the segmented original image to obtain the transformed coordinate set of the scattered block. This transformation can be achieved by converting the coordinates of the second scattered block from the image coordinate system of the segmented fused image to the image coordinate system of the segmented original image. Here, the coordinate transformation relationship between the image coordinate system of the segmented fused image and the image coordinate system of the segmented original image is the same as the transformation relationship between the image coordinate system of the original image and the image coordinate system of the fused image.

[0061] Step S5: Based on the first set of scattered block coordinates, the transformed set of scattered block coordinates is filtered to generate a target set of scattered block coordinates. The target set of scattered block coordinates represents the location distribution of scattered blocks whose displacement distance exceeds a preset distance threshold. Specifically, for each transformed scattered block coordinate, it can be determined whether there exists a first set of scattered block coordinates with a displacement less than the preset distance threshold. If so, the transformed scattered block coordinates are deleted, and the remaining transformed scattered block coordinates are used as the target set of scattered block coordinates.

[0062] In practice, to determine the extent of a landslide, it is first necessary to determine whether the landslide was caused by a landslide. However, directly matching the converted coordinates of the landslide with the coordinates of the first landslide can lead to significant matching errors due to the irregular shapes of the debris (e.g., rocks), and also consumes considerable computational resources due to the large number of coordinates. Therefore, the above implementation method eliminates the need for matching the converted coordinates of the landslide with the coordinates of the first landslide; instead, it determines whether the landslide's position has changed solely by the distance between the coordinates. Specifically, if rocks are densely distributed in a certain area (i.e., a large number of landslides caused by a landslide), the number of falsely identified deletions will also be relatively large. In this case, using the area where the original rocks were located as the coverage area can more stably reflect the overall distribution of the landslide extent. Conversely, if rocks are sparsely distributed (a small number of landslides appearing occasionally), the number of falsely identified deletions will be extremely small, and the impact on the final area division error will be negligible. Therefore, even if scattered blocks are mistakenly deleted due to coupling factors, there is no need for precise matching and correction of each block. The robustness of the result can be guaranteed by directly defining the sliding range based on the original block position.

[0063] Step S6: Landslide axis identification is performed on the segmented and fused image to generate an axial curve equation, where the axial curve equation represents the landslide axis of the landslide area. Here, the axial curve equation in the segmented and fused image can be determined using the skeleton algorithm in the OpenCV (Open Source Computer Vision Library) cross-platform computer vision library.

[0064] Step S7: Based on the aforementioned axial curve equation, landslide tongue detection is performed on the segmented and fused image to generate a landslide tongue boundary coordinate sequence and a landslide tongue confidence score. A preset detection model can be used to detect landslide tongues in the segmented and fused image, outputting a landslide tongue probability map and a binary segmentation mask. Here, the boundary coordinates in the binary segmentation mask can be determined as the landslide tongue boundary coordinates. Secondly, the average probability value corresponding to each landslide tongue boundary coordinate in the landslide tongue probability map can be determined as the landslide tongue confidence score. In practice, although landslide tongues are difficult to identify due to factors such as gravel accumulation and terrain modification, they are crucial structural units at the landslide leading edge, playing a vital role in defining the landslide range, determining the main sliding direction, evaluating stability, and delineating risk zones. Therefore, landslide tongue detection can be used to initially locate the boundary between the landslide and the original ground surface for subsequent landslide coverage location determination.

[0065] Specifically, the detection model can employ the BiSeNet V2 (Bilateral Segmentation Network, a bilateral segmentation network for real-time semantic segmentation) model. The input is a 5-channel tensor, and the output is the landslide tongue boundary coordinate sequence and landslide tongue confidence score. First, a landslide axial distance feature map can be generated based on the axial curve equation. Specifically, the pixel values ​​of coordinates on the axial curve equation in the segmented fused image can be set to zero to obtain the initial feature map. Then, the pixel values ​​of other coordinates are set according to their distance from the axial curve equation to obtain the filled feature map. Here, the farther the coordinate is from the axial curve equation, the larger the corresponding pixel value. Finally, the filled feature map is normalized to obtain the landslide axial distance feature map. Second, the landslide axial distance feature map and the corresponding spectral layers (i.e., red band, green band, blue band, and near-infrared band) of the segmented fused image can be used as a 5-channel tensor as input to the detection model, outputting a landslide tongue probability map and a binary segmentation mask.

[0066] Step S8: In response to the landslide tongue confidence level being greater than or equal to a preset edge confidence threshold, the landslide tongue boundary coordinate sequence is expanded according to the target scattered block coordinate set to generate an expanded boundary coordinate sequence. The landslide tongue confidence level being greater than or equal to the preset edge confidence threshold indicates that the landslide tongue recognition accuracy meets the recognition requirements. Next, the smallest bounding circle of each cluster coordinate corresponding to each target scattered block coordinate can be used as the scattered block region. Then, the smallest enclosing circle between each expanded boundary coordinate in the expanded boundary coordinate sequence and each initial coordinate in the initial coordinate sequence can be determined as the two-dimensional landslide area. Finally, boundary coordinates from the boundary coordinates of the two-dimensional landslide area whose distance from the initial coordinates is greater than a preset spacing threshold can be selected as expanded boundary coordinates to obtain the expanded boundary coordinate sequence. The expanded boundary coordinate sequence represents the landslide boundary after the landslide tongue boundary expansion based on the target scattered blocks.

[0067] Optionally, if the target debris coordinate set is empty after target screening, it indicates that there is a large amount of debris in the landslide area (i.e., the rocks in the landslide overlap significantly with the rocks at their original locations). Therefore, the coordinates of each transformed debris block can be used as the target debris block coordinates to expand the landslide tongue boundary coordinate sequence. This results in a larger coverage area corresponding to the expanded boundary coordinate sequence.

[0068] Step S9: Transform the expanded boundary coordinates in the above expanded boundary coordinate sequence from the image coordinate system to the planar geographic coordinate system to obtain the current boundary coordinate sequence. Specifically, the expanded boundary coordinates are transformed from the image coordinate system of the segmented and fused image to the planar geographic coordinate system through coordinate transformation to obtain the current boundary coordinates.

[0069] Step S10: Based on the aforementioned calibrated elevation dataset, determine the vertical coordinate values ​​of each current boundary coordinate in the current boundary coordinate sequence to generate a three-dimensional boundary coordinate sequence, which serves as the second edge coordinate sequence. Specifically, the calibrated elevation data in the calibrated elevation dataset that corresponds to the same position as the current boundary coordinate can be used as the vertical coordinate value of that current boundary coordinate, thereby generating the three-dimensional boundary coordinates. Finally, the three-dimensional boundary coordinates are used as the second edge coordinates.

[0070] Optionally, the aforementioned implementing entity may also include:

[0071] Step S1: In response to the landslide tongue confidence level being less than a preset edge confidence threshold, two first edge coordinates satisfying preset position conditions are selected from the first edge coordinate sequence as the first base point coordinates and the second base point coordinates. The preset position conditions are that the distance between the two first edge coordinates is maximized, and the vertical coordinate values ​​of the two first edge coordinates are less than the target vertical coordinate value. The first base point coordinates and the second base point coordinates respectively represent the boundary points of abrupt changes in landslide curvature. The target vertical coordinate value is the sum of the minimum vertical coordinate value of each first edge coordinate in the first edge coordinate sequence and a preset height value.

[0072] Step S2: Based on the first base point coordinates, the second base point coordinates, and the target scattered block coordinate set, a two-dimensional edge coordinate sequence is generated. First, the first and second base point coordinates are projected from the planar geographic coordinate system to the image coordinate system corresponding to the calibrated image, obtaining first and second projected coordinates. Then, the target scattered block coordinates with the largest distance between them and the straight lines containing the first and second base point coordinates are selected from the target scattered block coordinate set as the maximum distance scattered coordinates. Next, the minimum bounding sphere of each target scattered block coordinate in the first, second, and target scattered block coordinate sets is determined. The minimum bounding sphere is then divided into regions along the straight lines containing the first and second projected coordinates, resulting in a first region and a second region. Finally, the boundary coordinates of the region containing the maximum distance scattered coordinates (either the first or second region) are determined as two-dimensional edge coordinates, resulting in the two-dimensional edge coordinate sequence.

[0073] Step S3: Based on the above-mentioned calibrated elevation dataset and the above-mentioned two-dimensional edge coordinate sequence, a second edge coordinate sequence is generated. First, calibrated elevation data corresponding to the two-dimensional edge coordinates can be selected from the calibrated elevation dataset as elevation values. Then, the two-dimensional edge coordinates can be transformed from the image coordinate system corresponding to the segmented original image to the planar geographic coordinate system, and the elevation values ​​corresponding to the two-dimensional edge coordinates can be used as vertical coordinates to obtain three-dimensional scattered coordinates, which serve as the second edge coordinates.

[0074] As an example, see Figure 3 By identifying the scattered blocks 301 in the landslide area, the edges 303 of the difficult-to-define landslide tongue 302 can be expanded. This transforms the problem of identifying the difficult-to-define landslide tongue boundary into the problem of delineating the landslide impact area, thus facilitating the determination of the boundary (i.e., the partial boundary of the smallest enclosing circle) 304 of the landslide impact area based on each scattered block 301. Furthermore, this facilitates the provision of more reliable reference data for subsequent landslide disaster risk warnings.

[0075] Step 105: Perform landslide area simulation processing based on the original image, fused image, first edge coordinate sequence and second edge coordinate sequence to generate landslide surface area and landslide coverage area.

[0076] In some embodiments, the execution entity may perform landslide area simulation processing based on the original image, the fused image, the first edge coordinate sequence, and the second edge coordinate sequence to generate the landslide surface area and the landslide coverage area. The landslide surface area represents the exposed area of ​​the landslide on the ground surface. The landslide coverage area represents the area of ​​the landslide covering the ground surface. Here, both the landslide surface area and the landslide coverage area are estimated values.

[0077] Specifically, firstly, the geometric center of the landslide area enclosed by the first and second edge coordinate sequences is taken as the centroid. Then, the landslide axis is translated to the location of the centroid, and the corresponding elevation values ​​are determined. Here, the slope value corresponding to the centroid is calculated using these elevation values. Next, the first edge coordinates from the first edge coordinate sequence and the second edge coordinates from the aforementioned second edge coordinate sequence are projected onto the original image to obtain the projected edge coordinate sequence. The two-dimensional area of ​​the region enclosed by the projected edge coordinates in the aforementioned projected edge coordinate sequence is determined. Finally, the landslide surface area is generated using the pre-constructed inversion model: Landslide surface area = Two-dimensional area / cos(slope value × π / 180°). Furthermore, the aforementioned two-dimensional area can be used as the landslide coverage area.

[0078] In practice, by using the slope value corresponding to the centroid of the landslide as an engineering representative parameter of the overall landslide slope, a quantitative inversion relationship between the two-dimensional (projected) area and the actual slope area is established. This achieves a technical improvement in landslide area calculation, shifting from planar geometric statistics to topographic physical constraint correction. Consequently, without increasing additional data acquisition costs, the problem of the generally smaller two-dimensional projected area of ​​landslides in areas with higher slopes can be effectively corrected, significantly improving the accuracy of landslide area estimation results.

[0079] In some optional implementations of certain embodiments, the execution entity performs landslide area simulation processing based on the original image, the fused image, the first edge coordinate sequence, and the second edge coordinate sequence to generate the landslide surface area and the landslide coverage area, including:

[0080] Step S1 involves detecting engineering facilities in the original image to generate an engineering facility detection information set. This can be achieved by semantically segmenting the original scene image using a pre-trained U-Net semantic segmentation model, resulting in a semantic segmentation result set. The semantic segmentation result set can include detection labels and detection regions. The detection labels represent the semantics of the detection regions. For example, the semantics of a detection region could be ground, building, river, or road.

[0081] Step S2: Based on the aforementioned original imagery, the aforementioned calibrated elevation dataset, and the aforementioned engineering facility inspection information set, a three-dimensional simulated surface model is constructed in a planar geographic coordinate system. First, a digital elevation model (DEM) is established according to the calibrated elevation data corresponding to each coordinate in the original imagery. Then, regions are marked in the DEM according to the inspection areas included in the engineering facility inspection information, resulting in the three-dimensional simulated surface model. Here, each inspection marker can be pre-set with a rendering color to distinguish the regions corresponding to different inspection markers. The three-dimensional simulated surface model represents the three-dimensional surface structure of the post-earthquake risk area.

[0082] Step S3: Based on the aforementioned three-dimensional simulated surface model, determine the coverage areas corresponding to the first edge coordinate sequence and the second edge coordinate sequence, which are then used as the landslide coverage area. The planar coordinate system (i.e., the XY coordinate system) of the three-dimensional simulated surface model is the projected coordinate system. Therefore, within the planar coordinate system of the three-dimensional simulated surface model, a closed polygon formed by combining each first edge coordinate in the first edge coordinate sequence and each second edge coordinate in the second edge coordinate sequence can be determined. Then, the geometric area of ​​the closed polygon is calculated as the landslide coverage area.

[0083] Step S4: Based on the fused image, the first edge coordinate sequence, the second edge coordinate sequence, and the three-dimensional simulated surface model, a landslide mask is constructed. The landslide mask is a binarized spatial data layer used to precisely define the extent of the landslide area. Next, the elevation values ​​of each coordinate position in the fused image are determined using measurement software (e.g., RealityCapture 3D modeling software) and used as the fused image elevation values. Here, the fused image elevation values ​​characterize the height of the landslide body. Then, according to the coordinate transformation relationship between the fused image and the planar coordinate system in the three-dimensional simulated surface model, the fused image elevation values ​​corresponding to the coordinates in the fused image are replaced with the corresponding coordinate positions in the three-dimensional simulated surface model to obtain the current three-dimensional model. Optionally, the current three-dimensional model corresponding to the landslide area can also be reconstructed using the fused image, the first edge coordinate sequence, and the second edge coordinate sequence.

[0084] Next, an initial raster image is constructed according to the spatial resolution of the current 3D model and the preset number of raster rows and columns. Then, if the confidence level of the landslide tongue is greater than or equal to the preset edge confidence threshold, the coordinates of each first edge and each landslide tongue boundary coordinate in the landslide tongue boundary coordinate sequence can be converted into pixel coordinates in the initial raster image as raster coordinates. Then, using a scanline algorithm, each raster coordinate in the initial raster image is binarized to obtain a binarized raster image, which serves as a landslide mask. In the binarized raster image, pixels within the closed polygonal region formed by the merged raster coordinates have a value of 1, while pixels with other coordinates have a value of 0. Furthermore, if the confidence level of the landslide tongue is less than the preset edge confidence threshold, it indicates that the landslide tongue boundary is unclear. Therefore, the second edge coordinates can be replaced with the landslide tongue boundary coordinates to generate a binarized raster image.

[0085] Step S5: Based on the landslide mask described above, generate the surface area of ​​the landslide body. Specifically, for each grid cell within a closed polygonal region of the landslide mask, determine the slope value and the cosine of the slope value. Then, the ratio of the cell area of ​​the grid cell to the slope value is used as the grid area. Finally, the sum of the areas of all grid cells is determined as the surface area of ​​the landslide body.

[0086] In practice, this disclosure constructs a landslide mask through the above-described embodiments, and calculates the landslide coverage area and landslide surface area respectively, achieving comprehensive and high-precision quantification of the landslide's geometric characteristics. This allows for full utilization of a high-precision digital elevation model and slope correction methods, avoiding the area underestimation problem of traditional planar measurements, and improving the accuracy of the generated landslide surface area and landslide coverage area.

[0087] Optionally, the aforementioned implementing entity may also include:

[0088] Step S1 involves conducting a disaster chain risk assessment on the area corresponding to the landslide coverage area to generate disaster chain information. This disaster chain information indicates whether engineering facilities exist within the area corresponding to the landslide coverage area. Engineering facilities include buildings, roads, and water conservancy facilities. First, it is determined whether the landslide area corresponding to the landslide coverage area overlaps with the detection area representing engineering facilities. If they overlap, the landslide area in the post-earthquake risk zone is determined to pose a hazard to engineering facilities; therefore, disaster chain information indicating the presence of engineering facilities within the area corresponding to the landslide coverage area is generated.

[0089] Step S2: In response to the aforementioned disaster chain information meeting preset disaster conditions, a risk warning is issued to the target terminal. The preset disaster conditions are that the disaster chain information indicates the presence of engineering facilities within the area corresponding to the landslide coverage area. Furthermore, the target terminal can be a monitoring and early warning center platform.

[0090] As an example, if a landslide overlaps with the area where road facilities are located, it indicates that landslides often cause road blockages, and a risk warning of road blockage can be issued to the target terminal.

[0091] The above-described embodiments of this disclosure have the following beneficial effects: the landslide area simulation generation method for post-earthquake secondary disasters according to some embodiments of this disclosure can improve the accuracy of the generated landslide area. Specifically, the landslide area simulation generation method for post-earthquake secondary disasters according to some embodiments of this disclosure firstly acquires panchromatic and multispectral image sets, and obtains original images and elevation datasets. The panchromatic and multispectral images are high-resolution images acquired by a multi-view camera mounted on a UAV at a preset angle targeting the post-earthquake risk area. The original images are color images containing the aforementioned post-earthquake risk area, captured by an optical remote sensing satellite before the earthquake time point. The elevation dataset represents the height characteristics of the aforementioned post-earthquake risk area. Here, the panchromatic image possesses clear geometric texture features, facilitating more accurate extraction of the edge features of the landslide area. Compared to commonly used (red, green, and blue three-channel) color images, multispectral images can provide richer spectral information, allowing for better identification of the landslide area by combining the differences between the landslide area and the surrounding terrain. Furthermore, by introducing the original imagery and elevation dataset, the real topographic environment of the landslide area can be incorporated, thereby improving the accuracy of landslide area localization. Then, the panchromatic imagery is fused with the various multispectral images from the multispectral image set to generate a fused image. This fusion process allows the fused image to possess better texture detail and color features, thus improving the accuracy of landslide area localization. Next, landslide area identification is performed on the fused imagery to generate identification results. This landslide area identification can be used to further determine whether landslide hazards exist within the post-earthquake risk area. Subsequently, in response to the identification results indicating that the post-earthquake risk area contains landslide areas, landslide area edge detection is performed based on the original imagery, the elevation dataset, and the fused imagery to generate a first edge coordinate sequence and a second edge coordinate sequence, where the first and second edge coordinate sequences represent the landslide perimeter of the landslide area. In practice, the accuracy of landslide area identification depends on the precision of landslide boundary positioning. However, because the boundary line of the landslide toe (landslide tongue) is difficult to define, directly identifying the overall boundary of the landslide area can easily lead to significant errors. In contrast, the other locations within the landslide perimeter, except for the toe, have clearly defined surface features due to mountain displacement, making them easier to identify. Therefore, the boundary line of the landslide toe is decoupled from other boundaries, generating a first edge coordinate sequence and a second edge coordinate sequence. This reduces the impact of identifying the toe boundary line on the landslide perimeter identification, thereby improving the positioning accuracy of the landslide boundary. Finally, landslide area simulation processing is performed based on the original image, the fused image, the first edge coordinate sequence, and the second edge coordinate sequence to generate the landslide surface area and landslide coverage area.Here, by combining the real terrain environment (original image) and the landslide area boundary (first edge coordinate sequence and second edge coordinate sequence), a more accurate landslide surface area and landslide coverage area can be generated.

[0092] Further reference Figure 4 As an implementation of the methods shown in the above figures, this disclosure provides some embodiments of a landslide area simulation generation device for post-earthquake secondary disasters. These device embodiments are similar to... Figure 1 Corresponding to the method embodiments shown, the landslide area simulation generation device for post-earthquake secondary disasters can be specifically applied to various electronic devices.

[0093] like Figure 4 As shown, a landslide area simulation generation device 400 for post-earthquake secondary disasters in some embodiments includes: an acquisition and processing unit 401, a fusion processing unit 402, a landslide area identification unit 403, a landslide area edge detection unit 404, and a landslide area simulation processing unit 405. The acquisition and processing unit 401 is configured to acquire panchromatic images and multispectral image sets, and to acquire raw images and elevation datasets. The panchromatic images and multispectral images are high-resolution images acquired by a multi-view camera mounted on a UAV at a preset angle targeting the post-earthquake risk area. The raw images are color images containing the aforementioned post-earthquake risk area captured by an optical remote sensing satellite before the earthquake time point. The elevation dataset represents the height characteristics of the aforementioned post-earthquake risk area. The fusion processing unit 402 is configured to fuse the panchromatic images with the various multispectral images in the multispectral image set to generate a fused image. The landslide area identification unit 403 is configured to process the fused images... The image is combined to identify landslide areas and generate identification results. The landslide area edge detection unit 404 is configured to perform landslide area edge detection based on the original image, the elevation dataset, and the fused image in response to the identification results indicating that the post-earthquake risk area contains landslide areas, to generate a first edge coordinate sequence and a second edge coordinate sequence, wherein the first edge coordinate sequence and the second edge coordinate sequence represent the landslide perimeter of the landslide area. The landslide area simulation processing unit 405 is configured to perform landslide area simulation processing based on the original image, the fused image, the first edge coordinate sequence, and the second edge coordinate sequence to generate the landslide surface area and the landslide coverage area.

[0094] It is understandable that the units and references recorded in the landslide area simulation generation device 400 for post-earthquake secondary disasters are... Figure 1The steps in the described method correspond accordingly. Therefore, the operations, features, and beneficial effects described above for the method also apply to the landslide area simulation generation device 400 for post-earthquake secondary disasters and the units contained therein, and will not be repeated here.

[0095] The following is for reference. Figure 5 It shows a schematic diagram of the structure of an electronic device 500 (e.g., a computing device) suitable for implementing some embodiments of the present disclosure. Figure 5 The electronic device shown is merely an example and should not be construed as limiting the functionality and scope of the embodiments of this disclosure.

[0096] like Figure 5 As shown, the electronic device 500 may include a processing unit 501 (e.g., a central processing unit, a graphics processing unit, etc.), which can perform various appropriate actions and processes according to a program stored in the read-only memory 502 or a program loaded from the storage device 508 into the random access memory 503. The random access memory 503 also stores various programs and data required for the operation of the electronic device 500. The processing unit 501, the read-only memory 502, and the random access memory 503 are interconnected via a bus 504. An input / output interface 505 is also connected to the bus 504.

[0097] Typically, the following devices can be connected to the input / output interface 505: input devices 506 including, for example, a touchscreen, touchpad, keyboard, mouse, camera, microphone, accelerometer, gyroscope, etc.; output devices 507 including, for example, a liquid crystal display (LCD), speaker, vibrator, etc.; storage devices 508 including, for example, magnetic tape, hard disk, etc.; and communication devices 509. Communication device 509 allows electronic device 500 to communicate wirelessly or wiredly with other devices to exchange data. Although Figure 5 An electronic device 500 with various devices is shown; however, it should be understood that it is not required to implement or possess all of the devices shown. More or fewer devices may be implemented or possessed alternatively. Figure 5 Each box shown can represent a device or multiple devices as needed.

[0098] In particular, according to some embodiments of this disclosure, the processes described above with reference to the flowcharts can be implemented as computer software programs. For example, some embodiments of this disclosure include a computer program product comprising a computer program carried on a computer-readable medium, the computer program containing program code for performing the methods shown in the flowcharts. In such embodiments, the computer program can be downloaded and installed from a network via a communication device 509, or installed from a storage device 508, or installed from a read-only memory 502. When the computer program is executed by the processing device 501, it performs the functions defined above in the methods of some embodiments of this disclosure.

[0099] It should be noted that, in some embodiments of this disclosure, the computer-readable medium may be a computer-readable signal medium or a computer-readable storage medium, or any combination thereof. A computer-readable storage medium may be, for example,—but not limited to—an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor system, apparatus, or device, or any combination thereof. More specific examples of a computer-readable storage medium may include, but are not limited to: an electrical connection having one or more wires, a portable computer disk, a hard disk, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), optical fiber, portable compact disk read-only memory (CD-ROM), optical storage device, magnetic storage device, or any suitable combination thereof. In some embodiments of this disclosure, a computer-readable storage medium may be any tangible medium containing or storing a program that can be used by or in conjunction with an instruction execution system, apparatus, or device. In some embodiments of this disclosure, a computer-readable signal medium may include a data signal propagated in baseband or as part of a carrier wave, carrying computer-readable program code. Such propagated data signals may take various forms, including but not limited to electromagnetic signals, optical signals, or any suitable combination thereof. A computer-readable signal medium can be any computer-readable medium other than a computer-readable storage medium, which can send, propagate, or transmit a program for use by or in connection with an instruction execution system, apparatus, or device. The program code contained on the computer-readable medium can be transmitted using any suitable medium, including but not limited to: wires, optical fibers, RF (radio frequency), etc., or any suitable combination thereof.

[0100] In some implementations, clients and servers can communicate using any currently known or future-developed network protocol such as HTTP (Hypertext Transfer Protocol) and can interconnect with digital data communication (e.g., communication networks) of any form or medium. Examples of communication networks include local area networks (“LANs”), wide area networks (“WANs”), the Internet (e.g., the Internet of Things), and peer-to-peer networks (e.g., ad hoc peer-to-peer networks), as well as any currently known or future-developed networks.

[0101] The aforementioned computer-readable medium may be included in the aforementioned electronic device; or it may exist independently and not assembled into the electronic device. The aforementioned computer-readable medium carries one or more programs that, when executed by the electronic device, cause the electronic device to: acquire panchromatic and multispectral image sets, and acquire raw images and elevation datasets, wherein the panchromatic and multispectral images are high-resolution images acquired by a multi-view camera mounted on a UAV at a preset angle targeting the post-earthquake risk area; the raw images are color images containing the aforementioned post-earthquake risk area taken by an optical remote sensing satellite before the earthquake time point; and the elevation dataset represents the height characteristics of the aforementioned post-earthquake risk area; and the panchromatic images are compared with the various multispectral images in the aforementioned multispectral image set. A fusion process is performed to generate a fused image; landslide areas are identified in the fused image to generate identification results; in response to the identification results indicating that the post-earthquake risk area includes landslide areas, landslide area edge detection is performed based on the original image, the elevation dataset, and the fused image to generate a first edge coordinate sequence and a second edge coordinate sequence, wherein the first edge coordinate sequence and the second edge coordinate sequence represent the landslide perimeter of the landslide area; landslide area simulation processing is performed based on the original image, the fused image, the first edge coordinate sequence, and the second edge coordinate sequence to generate the landslide surface area and the landslide coverage area.

[0102] Computer program code for performing operations of some embodiments of this disclosure can be written in one or more programming languages ​​or a combination thereof, including object-oriented programming languages ​​such as Java, Smalltalk, and C++, and conventional procedural programming languages ​​such as the "C" language or similar programming languages. The program code can be executed entirely on the user's computer, partially on the user's computer, as a standalone software package, partially on the user's computer and partially on a remote computer, or entirely on a remote computer or server. In cases involving remote computers, the remote computer can be connected to the user's computer via any type of network—including a local area network (LAN) or a wide area network (WAN)—or can be connected to an external computer (e.g., via the Internet using an Internet service provider).

[0103] The flowcharts and block diagrams in the accompanying drawings illustrate the architecture, functionality, and operation of possible implementations of systems, methods, and computer program products according to various embodiments of this disclosure. In this regard, each block in a flowchart or block diagram may represent a module, segment, or portion of code containing one or more executable instructions for implementing a specified logical function. It should also be noted that in some alternative implementations, the functions indicated in the blocks may occur in a different order than those indicated in the drawings. For example, two consecutively indicated blocks may actually be executed substantially in parallel, and they may sometimes be executed in reverse order, depending on the functions involved. It should also be noted that each block in the block diagrams and / or flowcharts, and combinations of blocks in the block diagrams and / or flowcharts, can be implemented using a dedicated hardware-based system that performs the specified function or operation, or using a combination of dedicated hardware and computer instructions.

[0104] The functions described above in this document can be performed at least in part by one or more hardware logic components. For example, exemplary types of hardware logic components that can be used, without limitation, include: field-programmable gate arrays (FPGAs), application-specific integrated circuits (ASICs), application-specific standard products (ASSPs), system-on-a-chip (SoCs), complex programmable logic devices (CPLDs), and so on.

[0105] The above description is merely a selection of preferred embodiments of this disclosure and an explanation of the technical principles employed. Those skilled in the art should understand that the scope of the invention involved in the embodiments of this disclosure is not limited to technical solutions formed by specific combinations of the above-described technical features, but should also cover other technical solutions formed by arbitrary combinations of the above-described technical features or their equivalents without departing from the above-described inventive concept. For example, technical solutions formed by substituting the above-described features with (but not limited to) technical features with similar functions disclosed in the embodiments of this disclosure.

Claims

1. A method for simulating and generating landslide area for post-earthquake secondary disasters, characterized in that, include: The system acquires panchromatic and multispectral image sets, as well as raw images and elevation datasets. The panchromatic and multispectral images are high-resolution images acquired by a multi-view camera mounted on a drone at a preset angle for the post-earthquake risk area. The raw images are color images of the post-earthquake risk area taken by an optical remote sensing satellite before the earthquake time point. The elevation dataset represents the height characteristics of the post-earthquake risk area. The panchromatic image is fused with each multispectral image in the multispectral image group to generate a fused image. Landslide areas are identified in the fused image to generate identification results; In response to the identification result indicating that the post-earthquake risk area includes a landslide area, landslide area edge detection is performed based on the original image, the elevation dataset, and the fused image to generate a first edge coordinate sequence and a second edge coordinate sequence, wherein the first edge coordinate sequence and the second edge coordinate sequence represent the landslide perimeter of the landslide area; The landslide area is simulated based on the original image, the fused image, the first edge coordinate sequence, and the second edge coordinate sequence to generate the landslide surface area and the landslide coverage area.

2. The method for simulating and generating landslide area for post-earthquake secondary disasters according to claim 1, characterized in that, The acquisition of the original imagery and elevation dataset includes: Historical image maps containing the post-earthquake risk area and meeting preset filtering conditions are selected from the historical image map library and used as target image maps. The preset filtering conditions are that the time point corresponding to the historical image map is before the earthquake time point and the time interval between the historical image map and the earthquake time point is the shortest. The area corresponding to the post-earthquake risk area is extracted from the target image map as the original image; Elevation datasets are obtained by selecting the elevation data corresponding to the original image from the elevation database.

3. The method for simulating and generating landslide area for post-earthquake secondary disasters according to claim 1, characterized in that, The step of performing landslide area edge detection based on the original image, the elevation dataset, and the fused image to generate a first edge coordinate sequence and a second edge coordinate sequence includes: The original image and elevation dataset are calibrated to obtain calibrated image and calibrated elevation dataset. Edge detection is performed on the fused image to generate an initial coordinate sequence; The initial coordinates in the initial coordinate sequence are transformed from the image coordinate system corresponding to the fused image to the image coordinate system corresponding to the calibrated image to obtain the transformed coordinate sequence. The transformed coordinates in the transformed coordinate sequence are transformed from the image coordinate system corresponding to the calibrated image to the planar geographic coordinate system to obtain the planar geographic coordinate sequence; For each planar geographic coordinate in the planar geographic coordinate sequence, the vertical coordinate value of the planar geographic coordinate system is determined based on the calibrated elevation dataset to obtain the first edge coordinates; Based on the obtained first edge coordinate sequence, generate the second edge coordinate sequence.

4. The method for simulating and generating landslide area for post-earthquake secondary disasters according to claim 3, characterized in that, The step of generating a second edge coordinate sequence based on the obtained first edge coordinate sequence includes: Based on the first edge coordinate sequence, image segmentation is performed on the calibrated image and the fused image respectively to obtain the segmented original image and the segmented fused image. Scattered blocks are identified in the segmented original image to generate a first set of scattered block coordinates, wherein the first set of scattered block coordinates represents the location distribution of surface scattered blocks before the earthquake time point. Scattered blocks are identified in the segmented and fused image to generate a second set of scattered block coordinates, wherein the second set of scattered block coordinates represents the location distribution of surface scattered blocks after the earthquake time point. The second set of scattered block coordinates is transformed into the image coordinate system corresponding to the segmented original image to obtain the transformed set of scattered block coordinates; Based on the first set of scattered block coordinates, the transformed set of scattered block coordinates is filtered to generate a target set of scattered block coordinates, wherein the target set of scattered block coordinates represents the location distribution of scattered blocks whose displacement distance exceeds a preset distance threshold. Landslide axis identification is performed on the segmented and fused image to generate an axial curve equation, wherein the axial curve equation represents the landslide axis of the landslide area; Based on the axial curve equation, landslide tongue detection is performed on the segmented and fused image to generate a landslide tongue boundary coordinate sequence and landslide tongue confidence level; In response to the landslide tongue confidence being greater than or equal to a preset edge confidence threshold, the boundary coordinate sequence of the landslide tongue is expanded according to the target scattered block coordinate set to generate an expanded boundary coordinate sequence. The expanded boundary coordinates in the expanded boundary coordinate sequence are transformed from the image coordinate system to the planar geographic coordinate system to obtain the current boundary coordinate sequence; Based on the calibrated elevation dataset, the vertical coordinate values ​​of each current boundary coordinate in the current boundary coordinate sequence are determined to generate a three-dimensional boundary coordinate sequence, which serves as the second edge coordinate sequence.

5. The method for simulating landslide area generation for post-earthquake secondary disasters according to claim 4, characterized in that, The method further includes: In response to the landslide tongue confidence being less than a preset edge confidence threshold, two first edge coordinates that satisfy preset position conditions are selected from the first edge coordinate sequence as the first base point coordinates and the second base point coordinates. The preset position conditions are that the distance between the two first edge coordinates is the maximum and the vertical coordinates of the two first edge coordinates are less than the target vertical coordinate value. A two-dimensional edge coordinate sequence is generated based on the coordinates of the first base point, the coordinates of the second base point, and the target scattered block coordinate set; A second edge coordinate sequence is generated based on the calibrated elevation dataset and the two-dimensional edge coordinate sequence.

6. The method for simulating landslide area generation for post-earthquake secondary disasters according to claim 5, characterized in that, The step of performing landslide area simulation processing based on the original image, the fused image, the first edge coordinate sequence, and the second edge coordinate sequence to generate the landslide surface area and landslide coverage area includes: The original image is subjected to engineering facility inspection to generate an engineering facility inspection information set; Based on the original image, the calibrated elevation dataset, and the engineering facility detection information set, a three-dimensional simulated surface model is constructed in a plane geographic coordinate system; Based on the three-dimensional simulated surface model, the coverage areas corresponding to the first edge coordinate sequence and the second edge coordinate sequence are determined as the landslide body coverage area; A landslide mask is constructed based on the fused image, the first edge coordinate sequence, the second edge coordinate sequence, and the three-dimensional simulated surface model. The surface area of ​​the landslide body is generated based on the landslide cover.

7. The method for simulating landslide area generation for post-earthquake secondary disasters according to claim 6, characterized in that, The method further includes: A disaster chain risk assessment is performed on the area corresponding to the landslide coverage area to generate disaster chain information, wherein the disaster chain information indicates whether there are engineering facilities in the area corresponding to the landslide coverage area. In response to the disaster chain information meeting preset disaster conditions, a risk warning is issued to the target terminal.

8. A landslide area simulation generation device for post-earthquake secondary disasters, characterized in that, include: The acquisition and acquisition unit is configured to acquire panchromatic and multispectral image sets, as well as acquire raw images and elevation datasets. The panchromatic and multispectral images are high-resolution images acquired by a multi-view camera mounted on a UAV at a preset angle for the post-earthquake risk area. The raw images are color images containing the post-earthquake risk area taken by an optical remote sensing satellite before the earthquake time point. The elevation dataset represents the height characteristics of the post-earthquake risk area. The fusion processing unit is configured to fuse the panchromatic image with each multispectral image in the multispectral image group to generate a fused image. A landslide area identification unit is configured to identify landslide areas in the fused image to generate identification results; The landslide area edge detection unit is configured to perform landslide area edge detection based on the original image, the elevation dataset, and the fused image in response to the identification result indicating that the post-earthquake risk area contains a landslide area, to generate a first edge coordinate sequence and a second edge coordinate sequence, wherein the first edge coordinate sequence and the second edge coordinate sequence represent the landslide perimeter of the landslide area. The landslide area simulation processing unit is configured to perform landslide area simulation processing based on the original image, the fused image, the first edge coordinate sequence, and the second edge coordinate sequence to generate the landslide surface area and the landslide coverage area.

9. An electronic device, characterized in that, include: One or more processors; A storage device on which one or more programs are stored; When the one or more programs are executed by the one or more processors, the one or more processors implement the method as described in any one of claims 1 to 7.

10. A computer-readable medium, characterized in that, It stores a computer program thereon, wherein the computer program, when executed by a processor, implements the method as described in any one of claims 1 to 7.

Images