A method, device, medium, and product for dynamic monitoring of trench wall expansion erosion.
By using the Channel-DeepLab model and photogrammetry technology, the problems of automation and accuracy in the morphological changes of erosion gully walls have been solved, enabling high spatiotemporal resolution monitoring of the morphological characteristics of erosion gullies and landslides, supporting research on the development mechanism of erosion gullies and prevention and control.
Patent Information
- Application Number
- CN202410418066.6
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- Filing Date
- 2024-04-08
- Publication Date
- 2026-11-17
- Estimated Expiration
- 2044-04-08
AI Technical Summary
Existing technologies have low levels of automation and accuracy in identifying the expansion morphology of erosion gully walls and the dynamic identification of landslides. They are highly subjective, have limited image data processing capabilities, are difficult to operate, and have low spatiotemporal resolution, resulting in significant uncertainty in gully erosion prediction.
A boundary recognition model for landslides and erosion gullies trained using the Channel-DeepLab model was developed. By acquiring time-series orthophotos, a large amount of image data was automatically processed to extract high spatiotemporal resolution morphological features of erosion gully walls and landslides. Photogrammetry and ArcGIS were then used for binarization and geometric information calculation.
It achieves high-precision, low-cost, and easy-to-operate capture of information on erosion gully walls and landslides, provides high spatiotemporal resolution morphological feature analysis, supports research on the development mechanism of erosion gullies and prevention and control, and reduces measurement errors and equipment operation difficulty.
Smart Images

Figure CN118314459B_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of soil erosion process monitoring technology, and in particular to a method, device, medium and product for dynamic monitoring of gully wall expansion erosion. Background Technology
[0002] Morphological parameters of gully erosion channels are typically measured using contact monitoring methods or by directly calculating morphological indices using measuring tapes or surface measuring instruments. Recent developments in non-contact monitoring techniques for studying gully erosion development include satellite remote sensing, RTK, and lidar for large-scale monitoring; stereo photogrammetry for medium-scale monitoring; and three-dimensional laser scanning for small-scale (especially laboratory-scale) monitoring. However, these methods suffer from low automation and accuracy in identifying gully wall expansion morphology and dynamic gully width, high subjectivity, limited image data volume and processing capabilities, difficult equipment operation, and low spatiotemporal resolution. This results in the inability to track the spatiotemporal variations of individual gully walls and landslides at high spatiotemporal resolution benchmarks, leading to significant uncertainty in gully erosion prediction. Summary of the Invention
[0003] The purpose of this invention is to provide a method, device, medium, and product for dynamic monitoring of gully wall expansion erosion, which features easy operation, low cost, high precision, and high level of automation in monitoring gully development.
[0004] To achieve the above objectives, the present invention provides the following solution:
[0005] A method for dynamic monitoring of trench wall expansion erosion includes:
[0006] Obtain the time series of orthophotos of the monitored area on the slope to be monitored;
[0007] The orthophoto time series is input into the landslide boundary recognition model to obtain the landslide segmentation image time series; the landslide boundary recognition model is obtained by training the Channel-DeepLab model using historical landslide boundary labeled images;
[0008] The orthophoto time series is input into the erosion gully wall boundary recognition model to obtain the erosion gully wall segmentation image time series; the erosion gully wall boundary recognition model is obtained by training the Channel-DeepLab model using historical erosion gully wall boundary annotation images;
[0009] Based on the time sequence of the landslide segmentation images, the spatiotemporal morphological characteristics of the landslide within the monitoring area on the slope to be monitored are determined;
[0010] Based on the time sequence of segmented images of erosion gully walls, the spatiotemporal morphological characteristics of erosion gully walls within the monitored area on the slope to be monitored are determined.
[0011] According to specific embodiments provided by the present invention, the present invention discloses the following technical effects:
[0012] This invention provides a method, device, medium, and product for dynamic monitoring of gully wall expansion erosion, offering technical support for research on the development mechanism of gullies, the spatiotemporal variation characteristics of gully morphology, and soil erosion and prevention. Utilizing photogrammetry, it can quickly capture a large amount of image information on gully development containing temporal information, resulting in large data volume, high image data accuracy, low cost, minimal experimental consumables, high automation, and small measurement errors. It can accurately capture information on gully walls and landslides. Simultaneously, a Channel-DeepLab (gully width recognition network model) network model is constructed for identifying gully walls and landslides, enabling automated batch processing of the large amount of image resources obtained from photogrammetry. This outputs gully wall and landslide morphological features with high spatiotemporal resolution, offering advantages such as ease of operation, low cost, high precision, and high automation. Attached Figure Description
[0013] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.
[0014] Figure 1 Flowchart of the dynamic monitoring method for gully wall expansion erosion;
[0015] Figure 2 Schematic diagram of the principle of dynamic monitoring method for gully wall expansion erosion;
[0016] Figure 3 This is a schematic diagram of an indoor simulation test according to a specific embodiment of this application;
[0017] Figure 4 This is an example of the image preprocessing result using an image from a photogrammetric image database.
[0018] Figure 5 Example image showing the results of geometric information extraction of trench walls and landslide bodies;
[0019] Figure 6 Example diagram for specifying the movement trajectory of the collapsed body. Detailed Implementation
[0020] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.
[0021] The purpose of this invention is to provide a method, device, medium, and product for dynamic monitoring of gully wall expansion erosion, which features easy operation, low cost, high precision, and high level of automation in monitoring gully development.
[0022] To make the above-mentioned objects, features and advantages of the present invention more apparent and understandable, the present invention will be further described in detail below with reference to the accompanying drawings and specific embodiments.
[0023] Example 1
[0024] like Figure 1 As shown, a method for dynamic monitoring of trench wall expansion erosion in this embodiment includes:
[0025] Step 101: Obtain the time series of orthophotos of the area to be monitored on the slope to be monitored.
[0026] Step 102: Input the orthophoto time series into the landslide boundary recognition model to obtain the landslide segmentation image time series. The landslide boundary recognition model is obtained by training the Channel-DeepLab model using historical landslide boundary labeled images.
[0027] Step 103: Input the orthophoto time series into the erosion gully wall boundary recognition model to obtain the erosion gully wall segmentation image time series. The erosion gully wall boundary recognition model is obtained by training the Channel-DeepLab model using historical erosion gully wall boundary annotation images.
[0028] Step 104: Based on the time sequence of the landslide segmentation images, determine the spatiotemporal morphological characteristics of the landslide within the monitoring area on the slope to be monitored.
[0029] Step 105: Based on the temporal sequence of the segmented images of the erosion gully walls, determine the spatiotemporal morphological characteristics of the erosion gully walls within the monitored area on the slope to be monitored.
[0030] Step 101 includes:
[0031] Step 1011: Obtain the time series of orthophotos of the slope to be monitored.
[0032] Step 1012: Determine the rectangular frame enclosed by multiple control points in each orthophoto in the time series of orthophotos of the slope to be monitored as the clipping frame corresponding to the orthophoto.
[0033] Step 1013: Cut the orthophoto time series of the slope to be monitored according to multiple clipping frames to obtain the orthophoto time series of the area to be monitored on the slope.
[0034] Before step 101, the following is also included:
[0035] Step 106: Obtain multiple historical orthophotos of the slope to be monitored.
[0036] Step 107: Label the collapsed bodies in each historical orthophoto to obtain multiple historical collapsed body boundary label images.
[0037] Step 108: Determine the segmented image of the historical landslide body based on the boundary annotation image of the historical landslide body.
[0038] Step 109: Using historical orthophotos as input and historical landslide segmentation images as output, train the Channel-DeepLab model to obtain a landslide boundary recognition model.
[0039] Step 1010: Obtain multiple historical orthophotos of the slope to be monitored.
[0040] Step 1011: Label the erosion gully walls in each historical orthophoto to obtain multiple historical erosion gully wall boundary labeled images.
[0041] Step 1012: Determine the segmented image of the historical erosion gully wall based on the boundary annotation image of the historical erosion gully wall.
[0042] Step 1013: Using historical orthophotos as input and historical erosion gully wall segmentation images as output, train the Channel-DeepLab model to obtain an erosion gully wall boundary recognition model.
[0043] Step 104 includes:
[0044] Step 1041: Use ArcGIS 10.5 to binarize the time series of the landslide segmented image to obtain the time series of the landslide binarized image.
[0045] Step 1042: Construct an empty set as a spatiotemporal morphological feature set.
[0046] Step 1043: Let time number i = 1.
[0047] Step 1044: Determine the i-th image in the time sequence of the binarized images of the collapse body as the current binarized image of the collapse body.
[0048] Step 1045: Determine the time corresponding to the current binarized image of the collapsed body as the current time feature.
[0049] Step 1046: Obtain multiple closed regions in the current binarized image of the collapsed body as collapsed body regions.
[0050] Step 1047: Determine the area, perimeter, and center point coordinates of all collapsed areas as the current spatial morphological characteristics.
[0051] Step 1048: Determine the current time features and current spatial morphological features as the current spatiotemporal morphological features.
[0052] Step 1049: Add the current spatiotemporal morphological feature as the i-th element to the spatiotemporal morphological feature set, increment the value of time number i by 1, and return to step 1044 until the time sequence of the binarized image of the collapse body is traversed to obtain the spatiotemporal morphological features of the collapse body in the monitored area on the slope to be monitored.
[0053] Step 105 includes:
[0054] Step 1051: Use ArcGIS 10.5 to binarize the time series of segmented images of erosion gully walls to obtain the time series of binarized images of erosion gully walls.
[0055] Step 1052: Construct an empty set as a spatiotemporal morphological feature set.
[0056] Step 1053: Let time number i = 1.
[0057] Step 1054: Determine the i-th image in the time sequence of binarized images of the erosion trench wall as the current binarized image of the erosion trench wall.
[0058] Step 1055: Determine the time corresponding to the current binarized image of the erosion trench wall as the current time feature.
[0059] Step 1056: Obtain multiple closed regions in the current binarized image of the erosion trench wall as the erosion trench wall region.
[0060] Step 1057: Place multiple straight lines at equal intervals on the current binarized image of the erosion gully wall. All the straight lines are parallel to the Y-axis of the image coordinate system of the current binarized image of the erosion gully wall, and the X-axis of the image coordinate system is parallel to the projection direction of the slope line.
[0061] Step 1058: Determine any erosion gully wall area as the current erosion gully wall area.
[0062] Step 1059: Determine the area and perimeter of the current erosion gully wall region as the first spatial morphological feature.
[0063] Step 10510: Determine the straight line that intersects with the boundary of the current erosion gully wall area as the gully width line.
[0064] Step 10511: Determine any trench width line as the current trench width line.
[0065] Step 10512: Determine the horizontal coordinate of the current trench width line as the current trench width position.
[0066] Step 10513: Determine the absolute value of the difference between the ordinates of the two intersection points of the current gully width line and the boundary of the current erosion gully wall region as the gully width at the current gully width position.
[0067] Step 10514: Update the current trench width line and return to step 10511 until all trench width lines are traversed, and obtain the trench width of the current erosion trench wall area at different current trench width positions as the second spatial morphological feature.
[0068] Step 10515: Update the current erosion gully wall region and return to step 1059 until the erosion gully wall region is traversed to obtain the first spatial morphological features and second spatial morphological features of different erosion gully wall regions.
[0069] Step 10516: Determine the current time characteristics, and the first and second spatial morphological characteristics of different erosion gully wall areas as the current spatiotemporal morphological characteristics.
[0070] Step 10517: Add the current spatiotemporal morphological feature as the i-th element to the spatiotemporal morphological feature set, increment the value of time number i by 1, and return to step 1054 until the time sequence of the binarized image of the erosion gully wall is traversed to obtain the spatiotemporal morphological features of the erosion gully wall in the monitored area on the slope to be monitored.
[0071] The following is a detailed description of the dynamic monitoring method for trench wall expansion erosion provided in this embodiment, such as... Figure 2 This embodiment includes:
[0072] (1) Data acquisition of the expansion process of erosion trench walls with high spatiotemporal resolution.
[0073] like Figure 3 As shown, the initial experimental design included a ditch width of 0.1m, a depth of 0.04m, an upper runoff flow rate of 0.67L / s, and a slope of 5%. Downslope, middle slope, upslope, and transition sections were defined along the slope length, focusing on the ditch wall collapse and erosion process in the upslope, middle, and downslope sections. The experiment began when the upper runoff reached the outlet of the trench and ended when the ditch wall expanded to 0.05m from the edge of the trench. The duration of the experiment was divided into three segments: the early stage, the middle stage, and the late stage.
[0074] Photogrammetry was used to obtain the changes in the surface area and location of the gully wall expansion and landslide over time. Specifically, two Nikon D7000 (Nikon Inc., Melville, NY, USA) digital cameras were mounted on a rain-driving machine support at a distance of 3m from the soil surface at a slope length of 1.5m. Figure 3 As shown), ensure that the images captured by both cameras completely cover the test area. Adjust the mounting angle of the two cameras to make them parallel to the surface of the soil trough, and ensure that the overlap rate of the synchronously captured images is above 90%. To ensure normal and stable imaging and optimal field of view, the camera parameters need to be calibrated before the test to adapt to the laboratory lighting conditions. The final settings are f / 2.8 aperture, 800 ISO, 1 / 60 second shutter speed, and 20mm focal length. Ten control points are evenly distributed on both sides of the soil trough (e.g., ...). Figure 3 As shown in the figure, ensure that the plane where the control point is located is parallel to the soil surface, and use a total station to accurately measure the relative coordinates of the center of each control point to generate a high-precision orthophoto.
[0075] The two cameras are connected to the computer via USB cable. By presetting computer parameters, including automatic exposure time interval, image information storage path, image storage format and naming, the two cameras can be controlled to simultaneously and automatically perform image acquisition tasks and store the images in the specified path on the computer.
[0076] (2) Preprocessing of high-definition images.
[0077] ① Image selection and orthorectification. Images from the camera with clearer image quality and control points that can be automatically recognized by the program are selected as the data source (e.g., ...). Figure 4 As shown in section a), 10 control points are assigned measured coordinates, the projection transformation matrix is calculated, and the image is orthorectified to obtain an orthophoto. After resampling, the side length of one pixel in the image represents 1 mm in reality.
[0078] ② Image cropping. The intersection of the line connecting the upstream and downstream control points of the trench with the soil surface is taken as the upper and lower boundaries of the cropping. The left and right side walls of the trench are used as the left and right boundaries of the cropping. The image is then cropped to obtain the orthophoto of the experimental study area (e.g., ...). Figure 4 As shown in part b), the coordinates of the upper left corner of the cropped orthophoto are defined as (0, 0).
[0079] (3) Construct a Channel-DeepLab network model for identifying erosion gully walls and landslides.
[0080] ① Annotation of training and validation samples. Sixty typical images (preferably including various water boundaries and landslide types) were selected from the cropped orthophoto library. Forty images were used for model training, and 20 for model validation. Then, using the labelme toolkit on the Matlab platform, the labels were annotated on the walls of the erosion gullies (equivalent to the water surface within the erosion gully, such as…). Figure 4 (as shown in section f) and the collapsed body are labeled (e.g.) Figure 4 As shown in section d), it generates a annotation file that the program can learn.
[0081] ② Augmentation of Training and Validation Datasets. To increase the number of training and validation samples for deep learning, reduce computational load, and improve recognition accuracy, each cropped complete orthophoto (annotated file) (pixel value 2141×611) was further randomly cropped into 12 images of size 512×512 pixels, resulting in 720 annotated files. The square images after secondary cropping were then processed with rotation, flipping, brightness adjustments, and contrast adjustments to establish the training and validation datasets for deep learning.
[0082] ③ Channel-DeepLab Network Model Construction. The DeepLabV3+ network model was selected as the underlying framework, with the MobileNetV2 neural network as the backbone model. Semantic features were extracted to adapt to the processed images, and a preliminary Channel-DeepLab network model capable of automatically identifying the boundaries between water bodies and landslide bodies was constructed. Based on the Channel-DeepLab network model, boundary features of ditch wall boundary sample images and landslide body boundary sample images were extracted respectively. MobileNetV2 pre-trained weights were used with a frozen training method. The preliminary model was trained, evaluated, and corrected multiple times using the sample dataset until the prediction accuracy exceeded 90%, generating a ditch wall and landslide body boundary recognition network, and saving the weight file for ditch wall and landslide body boundary recognition. The model was used to batch process the pre-processed images from the erosion ditch image library, batch extracting the surfaces of ditch walls and landslide bodies, and outputting segmented images of ditch walls and landslide bodies (e.g., Figure 4 Part C and Figure 4 (Part e in the middle).
[0083] (4) Extraction of geometric information of erosion trench walls and collapse bodies with high spatiotemporal resolution.
[0084] ① Extraction of geometric information of the collapsed body: segmenting the image of the collapsed body (e.g.) Figure 5 Import the data (part a) into ArcGIS 10.5 and first perform binarization processing (e.g., ...). Figure 5 Part b and Figure 5 (part d in the middle), generate a vector file (such as...) Figure 5In the middle g part, fields such as sequence number, time, area, perimeter, and X and Y coordinates of the centroid are established, and geometric information is calculated and exported. This allows for the acquisition of morphological characteristics of the collapsed body with temporal and spatial information, which can be used for statistical analysis and for tracking the movement trajectory of a single collapsed body (e.g., ...). Figure 6 (As shown).
[0085] ② Extraction of geometric information of erosion trench walls: The segmented images of the trench walls obtained in the previous step (such as...) Figure 5 Import the middle part (c) into ArcGIS 10.5 and perform binarization processing (e.g., Figure 5 (Part d) Generate a vector file, create fields such as sequence number, time, area, and perimeter, calculate geometric information, and export.
[0086] ③ Extraction of gully width: In ArcGIS 10.5, a pair of vertical points is generated every 1 cm of slope length. A total of 213 pairs of points are generated for the entire experimental study area, corresponding to 213 gully widths (including transition sections). The distance between the two points in each pair must be greater than the width of the binarized segmented image. The "Point Set to Line" function is used to connect each pair of points into a line segment (e.g., ...). Figure 5 (Part e) Then, using the vector file of the water surface to trim line segments, the image at each moment can generate several line segments of varying lengths. The length of each line segment represents the width of the ditch at a certain slope at a certain moment (e.g., Figure 5 The middle f part and Figure 5 In the middle (h part), the geometric information of the line segment length is exported for subsequent analysis.
[0087] ④ Automatic output of equidistant trench widths. In this study, the width of the water surface is equal to the width of the trench, so the trench width can be extracted based on the binarized segmented image of the water body. The pixels (X1, Y1) on the left bank boundary of the water body are matched with the pixels (X1, Y2) on the right bank boundary at the same X coordinate to generate pixel pairs. The trench width is then equal to the difference in the Y coordinates of this pair × 1 mm. Finally, along the slope length direction, a corresponding trench width is output every 1 mm, thus obtaining the trench width feature changes with temporal and spatial information.
[0088] Example 2
[0089] A computer device includes: a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the steps of a method for dynamic monitoring of trench wall expansion erosion in Embodiment 1.
[0090] Example 3
[0091] A computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the steps of a method for dynamic monitoring of trench wall expansion erosion in Embodiment 1.
[0092] Example 4
[0093] A computer program product includes a computer program that, when executed by a processor, implements the steps of a method for dynamic monitoring of trench wall expansion erosion in Embodiment 1.
[0094] This document uses specific examples to illustrate the principles and implementation methods of the present invention. The descriptions of the above embodiments are only for the purpose of helping to understand the method and core ideas of the present invention. Furthermore, those skilled in the art will recognize that, based on the ideas of the present invention, there will be changes in the specific implementation methods and application scope. Therefore, the content of this specification should not be construed as a limitation of the present invention.
Claims
1. A method for dynamic monitoring of gully wall expansion erosion, characterized in that, include: Obtain the time series of orthophotos of the monitored area on the slope to be monitored; The orthophoto time series is input into the landslide boundary recognition model to obtain the landslide segmentation image time series; the landslide boundary recognition model is obtained by training the Channel-DeepLab model using historical landslide boundary labeled images; The orthophoto time series is input into the erosion gully wall boundary recognition model to obtain the erosion gully wall segmentation image time series; the erosion gully wall boundary recognition model is obtained by training the Channel-DeepLab model using historical erosion gully wall boundary annotation images; Based on the time sequence of the landslide segmentation images, the spatiotemporal morphological characteristics of the landslide within the monitoring area on the slope to be monitored are determined; Based on the time sequence of segmented images of erosion gully walls, the spatiotemporal morphological characteristics of erosion gully walls within the monitored area on the slope to be monitored are determined; Based on the time series of landslide segmentation images, the spatiotemporal morphological characteristics of the landslides within the monitored area on the slope to be monitored are determined, including: ArcGIS 10.5 was used to binarize the time series of segmented images of the landslide body to obtain the time series of binarized images of the landslide body. Construct an empty set as a spatiotemporal morphological feature set; Let time number i = 1; The i-th image in the time sequence of binarized images of the collapsed body is determined as the current binarized image of the collapsed body; The current time corresponding to the current binarized image of the collapsed body is determined as the current time feature; The multiple closed regions in the current binarized image of the collapsed body are identified as collapsed body regions. Determine the area, perimeter, and center point coordinates of all collapsed areas as the current spatial morphological characteristics; Determine the current temporal characteristics and current spatial morphological characteristics as the current spatiotemporal morphological characteristics; Add the current spatiotemporal morphological feature as the i-th element to the spatiotemporal morphological feature set, increment the value of time number i by 1, and return to the step "determine the i-th image in the time series of the binarized images of the collapse body as the current binarized image of the collapse body", until the time series of the binarized images of the collapse body is traversed to obtain the spatiotemporal morphological features of the collapse body in the area to be monitored on the slope to be monitored. Based on the time-series images of erosion gully walls, the spatiotemporal morphological characteristics of the erosion gully walls within the monitored area on the slope to be monitored are determined, including: The temporal sequence of segmented images of erosion gully walls was binarized using ArcGIS 10.5 to obtain the temporal sequence of binarized images of erosion gully walls. Construct an empty set as a spatiotemporal morphological feature set; Let time number i = 1; The i-th image in the time sequence of binarized images of the erosion gully wall is determined as the current binarized image of the erosion gully wall; The current time feature is determined by identifying the time corresponding to the current binarized image of the erosion trench wall. Obtain multiple closed regions from the current binarized image of the erosion gully wall as the erosion gully wall region; Multiple straight lines are set at equal intervals on the current binarized image of the erosion gully wall; all the straight lines are parallel to the Y-axis of the image coordinate system of the current binarized image of the erosion gully wall, and the X-axis of the image coordinate system is parallel to the projection direction of the slope line. Define any erosion gully wall area as the current erosion gully wall area; The area and perimeter of the current erosion gully wall region are determined as the first spatial morphological feature; The straight line that intersects with the boundary of the current erosion gully wall area is the gully width line; Define any trench width line as the current trench width line; The x-coordinate of the current trench width line is determined as the current trench width position; The absolute value of the difference between the ordinates of the two intersection points of the current gully width line and the boundary of the current gully wall region is the gully width at the current gully width position. Update the current trench width line and return to the step "determine the horizontal coordinate of the current trench width line as the current trench width position" until all trench width lines are traversed, and the trench width of the current erosion trench wall area at different current trench width positions is obtained as the second spatial morphological feature; Update the current erosion gully wall area and return to the step "determine the area and perimeter of the current erosion gully wall area as the first spatial morphological feature" until the erosion gully wall areas are traversed to obtain the first spatial morphological feature and the second spatial morphological feature of different erosion gully wall areas. The current temporal characteristics, as well as the first and second spatial morphological characteristics of different erosion gully wall regions, are determined as the current spatiotemporal morphological characteristics. Add the current spatiotemporal morphological feature as the i-th element to the spatiotemporal morphological feature set, increment the value of time number i by 1, and return to the step "determine the i-th image in the time series of binarized images of erosion gully wall as the current binarized image of erosion gully wall", until the time series of binarized images of erosion gully wall is traversed, and the spatiotemporal morphological features of erosion gully wall in the area to be monitored on the slope to be monitored are obtained.
2. The method for dynamic monitoring of trench wall expansion erosion according to claim 1, characterized in that, Obtain the time series of orthophotos of the monitored area on the slope to be monitored, including: Obtain the time series of orthophotos of the slope to be monitored; The rectangular frame enclosed by multiple control points in each orthophoto in the time series of orthophotos of the slope to be monitored is determined as the clipping frame corresponding to the orthophoto. The orthophoto time series of the slope to be monitored is obtained by cropping the orthophoto time series of the slope to be monitored using multiple clipping frames.
3. The method for dynamic monitoring of trench wall expansion erosion according to claim 1, characterized in that, Before acquiring the time-series sequence of orthophotos of the monitored area on the slope to be monitored, the following steps are also included: Acquire multiple historical orthophotos of the slope to be monitored; The collapsed bodies in each historical orthophoto were labeled to obtain multiple historical collapsed body boundary labeled images; Based on the historical landslide boundary-marked images, the segmented images of the historical landslides are determined; Using historical orthophotos as input and historical landslide segmentation images as output, the Channel-DeepLab model is trained to obtain a landslide boundary recognition model.
4. The method for dynamic monitoring of trench wall expansion erosion according to claim 1, characterized in that, Before acquiring the time-series sequence of orthophotos of the monitored area on the slope to be monitored, the following steps are also included: Acquire multiple historical orthophotos of the slope to be monitored; The walls of the erosion gullies in each historical orthophoto were labeled to obtain multiple images of the boundary labels of the historical erosion gullies. Based on the boundary-marked images of historical erosion gully walls, segmentation images of historical erosion gully walls are determined; Using historical orthophotos as input and segmented images of historical erosion gully walls as output, the Channel-DeepLab model is trained to obtain an erosion gully wall boundary recognition model.
5. A computer device, comprising: The memory and processor contain a computer program stored in the memory and executable on the processor, characterized in that the processor executes the computer program to implement a method for dynamic monitoring of trench wall expansion erosion according to any one of claims 1-4.
6. A computer-readable storage medium having a computer program stored thereon, characterized in that, When executed by a processor, the computer program implements a method for dynamic monitoring of trench wall expansion erosion as described in any one of claims 1-4.
7. A computer program product, comprising a computer program, characterized in that, When executed by a processor, the computer program implements a method for dynamic monitoring of trench wall expansion erosion as described in any one of claims 1-4.
