Search

CN-119399633-B - Ground subsidence monitoring method and system based on twin neural network fusion

CN119399633BCN 119399633 BCN119399633 BCN 119399633BCN-119399633-B

Abstract

The application relates to the technical field of ground collapse monitoring, in particular to a ground collapse monitoring method based on twin neural network fusion, which comprises the steps of obtaining a ground collapse monitoring range, and arranging control points and target points; the method comprises the steps of acquiring real-time InSAR scanning images, carrying out first pretreatment on the real-time InSAR scanning images to obtain InSAR collapse field images, acquiring a plurality of real-time 3D local scanning images, carrying out second pretreatment on the InSAR collapse field images, carrying out treatment on the InSAR partial scanning images through a twin neural network to obtain a plurality of 3D local collapse field images, formulating a weight rule, carrying out weight marking on the plurality of 3D local collapse field images according to the weight rule, then splicing the 3D local collapse field images with the weight marks to obtain 3D global collapse field images with the weight marks, carrying out splicing on the InSAR collapse field images and the 3D global collapse field images with the weight marks through the twin neural network to obtain InSAR seamless collapse field fusion images, and monitoring the InSAR seamless collapse field fusion images to obtain monitoring results. The method can make up the defect of a single technical means and realize the high-precision and full-coverage monitoring of the ground collapse.

Inventors

  • SHAO SHIWEI

Assignees

  • 武汉软件工程职业学院(武汉开放大学)

Dates

Publication Date
20260512
Application Date
20241111

Claims (10)

  1. 1. The ground subsidence monitoring method based on the twin neural network fusion is characterized by comprising the following steps of: acquiring a monitoring range of ground subsidence, and pre-arranging control points and target points in the monitoring range; Acquiring an InSAR scanning image in real time of the monitoring range, performing first preprocessing on the InSAR scanning image, and acquiring an InSAR collapse field image according to the InSAR scanning image subjected to the first preprocessing, wherein the InSAR scanning image is acquired through interferometric synthetic aperture radar InSAR; Acquiring a plurality of 3D local scanning images in real time of the monitoring range, performing second preprocessing on the plurality of 3D local scanning images, and processing the plurality of 3D local scanning images subjected to the second preprocessing through a twin neural network to acquire a plurality of 3D local collapse field images, wherein the 3D local scanning images are acquired through 3D laser scanning; Taking the control point and the target point as a first selected area, taking the InSAR collapsed field image and collapse obvious areas in a plurality of 3D local collapsed field images as a second selected area, wherein attribute data of the collapse obvious areas meet a preset collapse data threshold value, taking a collapse susceptible area as a third selected area, wherein the collapse susceptible area accords with a preset collapse risk attribute, taking a high-quality area in a plurality of 3D local collapsed field images as a fourth selected area, wherein attribute data of the high-quality area meets a preset data quality threshold value, setting weights for the first selected area, the second selected area, the third selected area and the fourth selected area respectively, respectively making four independent weight rules for the first selected area, the second selected area, the third selected area and the fourth selected area and the weights corresponding to the third selected area, traversing each 3D local field image, carrying out a calculation on the collapse local area of each 3D local area according to the weight rules, carrying out a calculation on the collapse local area with the weights to obtain a plurality of the 3D local collapse images, and carrying out a calculation on the 3D local collapse image with the total marks according to the weight rules to the obtained total collapse marks; And splicing the InSAR collapse field image and the 3D global collapse field image with the weight mark through the twin neural network to obtain an InSAR seamless collapse field fusion image, and monitoring the InSAR seamless collapse field fusion image to obtain a monitoring result of the monitoring range.
  2. 2. The ground subsidence monitoring method based on the twin neural network fusion according to claim 1, wherein the InSAR collapse field image is a holed InSAR collapse field image, wherein a holed region of the holed InSAR collapse field image is obtained by removing a plurality of error regions by using a masking method, and wherein attribute data of the error regions meet a preset error threshold.
  3. 3. The ground subsidence monitoring method based on the twin neural network fusion according to claim 1, wherein the weighting of the region in each of the 3D partial collapse field images according to the weighting rule specifically comprises: carrying out corresponding weight marking on the region of the 3D partial collapse field image meeting one rule in the weight rules; carrying out weight superposition marking on the areas of the 3D partial collapse field image meeting two or more rules in the weight rules; and setting the weights of other areas of the 3D partial collapse field image which do not meet the weight rule as initial weights.
  4. 4. A ground collapse monitoring method based on twin neural network fusion according to claim 3, wherein the initial weights are each smaller than the weights set in the weight rule.
  5. 5. The ground subsidence monitoring method based on the twin neural network fusion according to claim 1, wherein the splicing of the InSAR subsidence field image and the 3D global subsidence field image with the weight mark through the twin neural network to obtain an InSAR seamless subsidence field fusion image, and the monitoring of the InSAR seamless subsidence field fusion image to obtain the monitoring result of the monitoring range specifically comprises the following steps: Respectively acquiring phase information and amplitude information of the InSAR collapsed field image, and encoding the phase information and the amplitude information in different channels through an InSAR data encoder to obtain InSAR collapsed field encoding data, wherein the InSAR encoder adopts a convolutional neural network structure; Encoding the 3D global collapse field image with the weight marks through a 3D point cloud data encoder to obtain 3D global collapse field encoded data, wherein the 3D point cloud data encoder adopts a point cloud convolutional neural network structure; Inputting the 3D global collapse field coding data and the InSAR collapse field coding data into the twin neural network, extracting information of a hole area in the InSAR collapse field image from the InSAR collapse field coding data by the twin neural network, and splicing and filling the hole area by using the 3D global collapse field coding data and the information of the hole area to obtain an InSAR seamless collapse field image; the twin neural network extracts a region subjected to weight marking from the 3D global collapse field coding data as a selected region, acquires a corresponding selected region in the InSAR seamless collapse field image according to the selected region, and performs displacement fusion on the corresponding selected region in the InSAR seamless collapse field image by using the 3D global collapse field coding data to obtain an InSAR seamless collapse field fusion image; and monitoring the InSAR seamless collapse field fusion images at different moments to obtain monitoring results of the monitoring range.
  6. 6. A ground subsidence monitoring system based on a twin neural network fusion, comprising: The range acquisition module is used for acquiring a monitoring range of ground subsidence, and control points and target points are arranged in the monitoring range in advance; The InSAR data acquisition-processing module is used for acquiring an InSAR scanning image in real time in the monitoring range, carrying out first preprocessing on the InSAR scanning image, and acquiring an InSAR collapse field image according to the InSAR scanning image subjected to the first preprocessing, wherein the InSAR scanning image is acquired through an interferometric synthetic aperture radar InSAR; The 3D scanning local data acquisition-processing module is used for acquiring a plurality of 3D local scanning images in real time of the monitoring range, carrying out second preprocessing on the plurality of 3D local scanning images, and processing the plurality of 3D local scanning images subjected to the second preprocessing through a twin neural network to acquire a plurality of 3D local collapse field images, wherein the 3D local scanning images are acquired through 3D laser scanning; The 3D scanning global data acquisition module is used for taking the control point and the target point as a first selected area, taking a collapse obvious area in the InSAR collapse field image and a plurality of 3D local collapse field images as a second selected area, wherein attribute data of the collapse obvious area meets a preset collapse data threshold value, taking a collapse susceptible area as a third selected area, wherein the collapse susceptible area meets a preset collapse risk attribute, taking a high-quality area in a plurality of 3D local collapse field images as a fourth selected area, wherein attribute data of the high-quality area meets a preset data quality threshold value, respectively setting weights for the first selected area, the second selected area, the third selected area and the fourth selected area, respectively making four independent weight rules for the first selected area, the second selected area, the third selected area and the fourth selected area and the corresponding weights thereof, carrying out the computation on the collapse image of each 3D local collapse field image according to the weight rules, and carrying out the computation on the three-dimensional local collapse image with all the three-dimensional (3D) of the 3D local collapse field images by the weight rule, and obtaining a plurality of local collapse images with all three-dimensional (3D) of the three-dimensional local collapse image with all three-dimensional (3D) local collapse image with all the three-dimensional image with all three-dimensional image (3D) all three-dimensional image with all three-dimensional image data; The splicing monitoring module is used for splicing the InSAR collapse field image and the 3D global collapse field image with the weight mark through the twin neural network to obtain an InSAR seamless collapse field fusion image, monitoring the InSAR seamless collapse field fusion image and obtaining a monitoring result of the monitoring range.
  7. 7. The ground collapse monitoring system based on twin neural network fusion of claim 6, wherein the 3D scan global data acquisition module is further configured to: carrying out corresponding weight marking on the region of the 3D partial collapse field image meeting one rule in the weight rules; carrying out weight superposition marking on the areas of the 3D partial collapse field image meeting two or more rules in the weight rules; and setting the weights of other areas of the 3D partial collapse field image which do not meet the weight rule as initial weights.
  8. 8. The ground collapse monitoring system based on twin neural network fusion of claim 6, wherein the splice monitoring module is further configured to: Respectively acquiring phase information and amplitude information of the InSAR collapsed field image, and encoding the phase information and the amplitude information in different channels through an InSAR data encoder to obtain InSAR collapsed field encoding data, wherein the InSAR encoder adopts a convolutional neural network structure; Encoding the 3D global collapse field image with the weight marks through a 3D point cloud data encoder to obtain 3D global collapse field encoded data, wherein the 3D point cloud data encoder adopts a point cloud convolutional neural network structure; Inputting the 3D global collapse field coding data and the InSAR collapse field coding data into the twin neural network, extracting information of a hole area in the InSAR collapse field image from the InSAR collapse field coding data by the twin neural network, and splicing and filling the hole area by using the 3D global collapse field coding data and the information of the hole area to obtain an InSAR seamless collapse field image; the twin neural network extracts a region subjected to weight marking from the 3D global collapse field coding data as a selected region, acquires a corresponding selected region in the InSAR seamless collapse field image according to the selected region, and performs displacement fusion on the corresponding selected region in the InSAR seamless collapse field image by using the 3D global collapse field coding data to obtain an InSAR seamless collapse field fusion image; and monitoring the InSAR seamless collapse field fusion images at different moments to obtain monitoring results of the monitoring range.
  9. 9. An electronic device comprising a memory and a processor, wherein the memory stores computer readable instructions, and the processor executes the computer readable instructions to implement a ground collapse monitoring method based on twin neural network fusion as claimed in any one of claims 1-5.
  10. 10. A computer storage medium having a computer program stored thereon, wherein the computer program when executed implements a ground collapse monitoring method based on twin neural network fusion as claimed in any one of claims 1 to 5.

Description

Ground subsidence monitoring method and system based on twin neural network fusion Technical Field The invention relates to the technical field of ground collapse monitoring, in particular to a ground collapse monitoring method and system based on twin neural network fusion. Background Ground subsidence is a common geological disaster that is usually caused by natural factors such as over-mining of groundwater, seismic activity, karst action, etc., and also may be caused by human factors such as mining activity, tunnel construction, underground engineering, etc. This phenomenon can cause subsidence of the earth's surface, affect the stability of the building, destroy the infrastructure, and even endanger human life. The occurrence of ground collapse is often sudden and hidden, and brings great challenges to monitoring and early warning work. The existing ground subsidence monitoring method mainly depends on a single technical means, for example, an interferometric synthetic aperture radar InSAR technology can provide large-scale ground deformation data, but data loss or errors possibly exist, a 3D laser scanning technology can provide high-precision local ground deformation data, but the monitoring range is limited, and the existing technology also combines the two technologies, but only uses a simple splicing technology, so that higher precision and accurate monitoring cannot be achieved. The above methods have advantages and disadvantages, and cannot be used for single full coverage and high-precision monitoring of ground collapse. Therefore, there is a need for improvements in the ground collapse method. Disclosure of Invention The invention aims to overcome at least one defect (deficiency) of the prior art, and provides a ground collapse monitoring method and system based on twin neural network fusion, which realize high-precision and full-coverage monitoring of ground collapse. Acquiring a monitoring range of ground subsidence, and pre-arranging control points and target points in the monitoring range; Acquiring an InSAR scanning image in real time of the monitoring range, performing first preprocessing on the InSAR scanning image, and acquiring an InSAR collapse field image according to the InSAR scanning image subjected to the first preprocessing, wherein the InSAR scanning image is acquired through interferometric synthetic aperture radar InSAR; Acquiring a plurality of 3D local scanning images in real time of the monitoring range, performing second preprocessing on the plurality of 3D local scanning images, and processing the plurality of 3D local scanning images subjected to the second preprocessing through a twin neural network to acquire a plurality of 3D local collapse field images, wherein the 3D local scanning images are acquired through 3D laser scanning; A weight rule is formulated according to the control point, the target point, the InSAR collapse field image and the 3D local collapse field images, the 3D local collapse field images are subjected to weight marking according to the weight rule and then spliced, and a 3D global collapse field image with weight marks is obtained; And splicing the InSAR collapse field image and the 3D global collapse field image with the weight mark through the twin neural network to obtain an InSAR seamless collapse field fusion image, and monitoring the InSAR seamless collapse field fusion image to obtain a monitoring result of the monitoring range. It can be understood that, because in the present technology, the ground collapse is monitored mainly by using the interferometric synthetic aperture radar InSAR and the 3D laser scanning technology, but the ground collapse is monitored by using the interferometric synthetic aperture radar InSAR technology alone, a large range of terrains can be monitored, but serious data errors occur in some areas, for example, the ground collapse is monitored by using the 3D laser scanning technology alone, although high-precision scanning data can be obtained, only a small range of terrains can be scanned at a time, complex and large splicing work is needed later, and large superposition errors occur at the boundary of the splicing work. Therefore, the method combines the interferometric synthetic aperture radar InSAR and the 3D laser scanning technology to monitor the subsidence ground, uses the twin neural network and the weight rule to splice, reduces errors caused by splicing, accurately monitors important areas, ensures that important areas of formed subsidence field images all have relatively high-precision data, and has a comprehensive coverage range. Optionally, the InSAR collapse field image is a holed InSAR collapse field image, wherein a holed area of the holed InSAR collapse field image is obtained by removing a plurality of error areas by using a masking method, and attribute data of the error areas meet a preset error threshold. It can be appreciated that the mask method is used for removing the image error reg