Slope monitoring and early warning method based on oblique photography and radar technology

Abstract

The application discloses a kind of slope monitoring early warning method based on oblique photography and radar technology, obtains the three-dimensional point cloud data and surface texture data of slope;Risk identification model is constructed;Three-dimensional point cloud data and surface texture data are input into registration fusion module and are registered and fused, obtain the three-dimensional image data of slope;The three-dimensional image data of slope is input into image recognition module and is cracked, obtains crack image, and crack image includes crack position, size and quantity;Crack image is input into comparison module, whether risk alarm is judged based on preset threshold value;Further, the scanning frequency of synthetic aperture radar is dynamically adjusted by crack image.The application can realize high-precision, comprehensive identification to slope crack by fusing the three-dimensional point cloud data obtained by multi-line laser radar and the surface texture data obtained by oblique photography unmanned aerial vehicle.

Description

Technical Field

[0001] The present invention belongs to the technical field of slope monitoring and early warning, and in particular relates to a slope monitoring and early warning method based on oblique photography and radar technology. Background Art

[0002] Slope monitoring is a crucial tool for preventing geological disasters and ensuring infrastructure safety. Traditional slope monitoring technologies rely primarily on single sensors or monitoring methods, such as surface displacement sensors, rain gauges, or total stations. While these methods can provide sufficient monitoring data, they have numerous limitations in complex environments. For example, surface monitoring equipment struggles to penetrate vegetation, making it impossible to obtain information from slope interiors or hidden areas. While satellite remote sensing technology has wide coverage, its resolution and data accuracy are significantly affected in areas with dense vegetation or complex terrain.

[0003] In recent years, the development of drone and radar technologies has significantly enriched slope monitoring methods. Drone oblique photography can quickly acquire high-resolution surface texture data of slopes, allowing for intuitive visualization of slope topography and crack distribution through 3D modeling. However, this technology can result in data loss in some areas when obscured by vegetation such as trees. Radar technology, particularly synthetic aperture radar (SAR) and geological radar (GPR), can penetrate vegetation and soil to obtain structural information within slopes. However, radar images have relatively low resolution and are unable to directly capture surface details.

[0004] The fusion of multi-source data obtained from oblique photography and radar technology is a key technical challenge in slope monitoring. While oblique photography provides high-resolution surface texture data that clearly reveals the location, size, and shape of cracks, it cannot penetrate vegetation obstructions. Meanwhile, while radar technology can penetrate vegetation obstructions and obtain internal structural information about the slope, its data resolution is low, making it difficult to directly identify cracks. Furthermore, the heterogeneity of multi-source data (e.g., differences in data format, dimension, and temporal resolution) poses challenges to data fusion.

[0005] Furthermore, in traditional slope monitoring, synthetic aperture radar (SAR) scanning and crack image data have long been separated: the SAR radar mechanically scans the slope partitions according to a fixed grid and cannot associate the crack risk characteristics, resulting in insufficient scanning of high-risk areas and wasted resources in low-risk areas. Summary of the Invention

[0006] In view of this, the present invention provides a slope monitoring and early warning method based on oblique photography and radar technology. Through multi-source data fusion, dynamic threshold classification and intelligent prediction algorithm, it can realize real-time monitoring of all elements of high and steep slopes and accurate early warning of disaster risks.

[0007] To achieve the above object, the technical solution of the present invention is achieved as follows:

[0008] A slope monitoring and early warning method based on oblique photography and radar technology, comprising:

[0009] S1: Acquire 3D point cloud data of the slope using multi-line laser radar; acquire surface texture data of the slope using oblique photography drones; surface texture data includes the location, size, and number of cracks on the slope surface;

[0010] S2: Build a risk identification model, which includes: registration and fusion module, image recognition module and comparison module;

[0011] Input the 3D point cloud data and surface texture data into the registration and fusion module for registration and fusion to obtain the 3D image data of the slope;

[0012] Inputting the three-dimensional image data of the slope into the image recognition module to perform crack recognition and obtain a crack image, which includes the crack location, size and number;

[0013] The crack image is input into the comparison module, and the risk type is determined based on the preset threshold value and a graded alarm is issued;

[0014] S3: Perform zonal scanning and detection of the slope using synthetic aperture radar (SAR). Based on the crack image in S2, the zoning is optimized to obtain the optimized detection area. The scanning frequency of the SAR is dynamically adjusted based on the risk type of the optimized detection area.

[0015] Among them, the optimized detection area is obtained as follows:

[0016] ;

[0017] in, Indicates the optimized detection area; Represents the position coordinates of the pixel points in the optimized detection area; Represents cracks in the crack image The center position coordinates of Represents any crack in the crack image The center position coordinates of represents the Euclidean distance; Represents the crack in the corresponding crack image Type of risk The risk scaling factor of Represents any crack in the corresponding crack image Type of risk The risk scaling factor, represents the risk scaling factor.

[0018] Furthermore, in step S1, the three-dimensional point cloud data of the slope is obtained by multi-line laser radar as follows:

[0019] Use multi-line laser radar to set horizontal angular resolution, vertical angular resolution and point cloud density; obtain 3D point cloud data of the slope through multi-station scanning and stitching , expressed as:

[0020] ;

[0021] in, Indicates the first points Axis coordinates, Indicates the first points Axis coordinates, Indicates the first points axis coordinates;

[0022] The surface texture data of the slope is obtained by using an oblique photography drone in step S1, specifically:

[0023] The oblique photography drone flies crosswise at a preset inclination angle to generate orthophotos. The crack texture of the orthophotos is enhanced using the HSV algorithm to obtain the surface texture data of the slope.

[0024] Furthermore, the three-dimensional image data of the slope is obtained in step S2, specifically:

[0025] Spatial registration, expressed as:

[0026] ;

[0027] in, represents the rotation matrix, represents the translation vector, Represents surface texture data;

[0028] By minimizing the distance error between corresponding points , solve the rotation matrix and translation vector to geometrically align the surface texture data and the 3D point cloud data;

[0029] The registered surface texture data and 3D point cloud data are fused to obtain the 3D image data of the slope.

[0030] Furthermore, in step S2, the three-dimensional image data of the slope is input into the image recognition module for crack recognition to obtain a crack image, specifically:

[0031] Step S21: preprocessing and data enhancement of the three-dimensional image data of the slope to obtain an enhanced two-dimensional grayscale image;

[0032] Step S22: performing multi-scale feature extraction, attention enhancement, and decoding on the enhanced two-dimensional grayscale image to obtain a crack mask image, i.e., a crack image; in this process, obtaining a count label of the crack mask image;

[0033] Step S23: performing a morphological closing operation on the crack mask image to optimize the crack boundary, and performing noise filtering based on a threshold of the connected domain area and the image size, to finally obtain a crack mask image after noise filtering;

[0034] Step S24: obtaining the position, size and number of cracks based on the crack mask image after noise filtering.

[0035] Furthermore, step S21 is specifically as follows:

[0036] Preprocess the input slope 3D image data and expand the 3D image data into a 2D grayscale image , to preserve the brightness difference between the crack and the image background, expressed as;

[0037] ;

[0038] in, are the red channel value, green channel value and blue channel value of the three-dimensional image data respectively, All are preset weight parameters;

[0039] pass Algorithm to enhance two-dimensional grayscale images The local contrast of the image is enhanced to obtain the enhanced two-dimensional grayscale image. , expressed as:

[0040] ;

[0041] in, Represents the contrast-constrained adaptive histogram equalization function.

[0042] Furthermore, step S22 is specifically as follows:

[0043] Enhanced 2D grayscale image Perform multi-scale feature extraction to obtain multi-scale features , expressed as: ;

[0044] in, represents a lightweight convolutional layer used to extract the local texture of the crack; represents the dilated spatial pyramid pooling operation, which is used to capture the features of both fine and wide cracks;

[0045] Attention enhancement is performed on the multi-scale features to obtain enhanced multi-scale features , denoted as:

[0046] ;

[0047] wherein, represents channel attention, used to suppress image interference;

[0048] The crack mask image is generated by the decoder , denoted as:

[0049] ;

[0050] wherein, represents the decoder, represents the Sigmoid activation function,

[0051] The crack mask image is separated to obtain a single crack image and is labeled, denoted as:

[0052] ;

[0053] wherein, represents a connected domain marking function, used to distinguish between adhered or crossed cracks, represents the count label of a single crack image, represents a preset threshold.

[0054] Further, step S23, specifically:

[0055] The crack mask image is morphologically filtered to obtain an optimized crack mask image, denoted as: ;

[0056] wherein, represents a closing operation; represents an expansion function, used to expand the boundary of the crack image; represents an erosion function, used to reduce the boundary of the crack image; represents the optimized crack mask image;

[0057] The optimized crack mask image is filtered to obtain a filtered crack mask image , denoted as:

[0058] ;

[0059] wherein, represents an indicator function, represents the connected domain area of ​​the input optimized crack mask image, represents the height of the input optimized crack mask image, represents the width of the input optimized crack mask image, is the preset noise figure.

[0060] Furthermore, step S24 is specifically as follows:

[0061] Get the crack mask image after noise filtering Location coordinates , expressed as:

[0062] ;

[0063] in, express The The crack coordinates of the pixel points, Represents the crack mask image after noise filtering Total number of pixels; Represents the connected domain area of ​​the input optimized crack mask image;

[0064] The crack mask image after noise filtering Perform size calculations to obtain the length of the crack , average width and maximum width , expressed as:

[0065] ;

[0066] ;

[0067] ;

[0068] in, Indicates the skeletonization of a single crack. A sequence of waypoints; Indicates the skeletonization of a single crack. A sequence of waypoints; represents the Euclidean distance; represents the ranging function;

[0069] The crack mask image after noise filtering Perform quantitative statistics to obtain the number of cracks , expressed as:

[0070] ;

[0071] in, Represents the maximum value function.

[0072] Further, the process of judging whether to perform risk warning based on the preset threshold value is specifically:

[0073] Obtain the risk coefficient , which is expressed as:

[0074] ;

[0075] Wherein, , and are preset weight parameters, represents the preset length of the crack, represents the preset maximum width of the crack, represents the preset maximum number of cracks;

[0076] Based on the risk coefficient Obtain the risk type To perform hierarchical alarm, specifically:

[0077] ;

[0078] Wherein, 0, 1 and 2 are labels corresponding to each risk type.

[0079] Further, the risk scaling factor in S3 corresponds to the risk type of the region, specifically:

[0080] When the risk type is a high-risk area, the risk scaling factor > 1;

[0081] When the risk type is a medium-risk area, the risk scaling factor = 1;

[0082] When the risk type is a low-risk area, the risk scaling factor < 1.

[0083] The beneficial effects of the present application are:

[0084] 1. By fusing the three-dimensional point cloud data obtained by the multi-line laser radar and the surface texture data obtained by the tilt photography unmanned aerial vehicle, high-precision and comprehensive identification of the slope crack can be realized.

[0085] The specific manifestations are as follows: accurate extraction of crack information: three-dimensional point cloud data provides accurate spatial geometric information of the slope, and can accurately locate the position and size of the crack; the surface texture data clearly shows the morphology and quantity of the crack through high-resolution images. The fusion of the two makes the crack identification no longer limited to a single dimension, but can accurately locate and quantify the crack characteristics in three-dimensional space; break through the limitation of shielding: although the oblique photography technology can obtain high-resolution surface texture, it has limitations when facing vegetation shielding such as trees. The three-dimensional point cloud data of the laser radar can penetrate part of the vegetation shielding and obtain the terrain information of the shielding area. Through data fusion, the missing crack information in the shielding area can be effectively filled, and the overall monitoring of the slope crack can be realized. Enhance the quality of model input: the three-dimensional image data obtained after the fusion of multi-source data is used as the input of the risk identification model, which provides more rich and accurate crack feature information for the model. This helps to improve the model's ability to identify and predict the crack risk, so as to more reliably evaluate the stability of the slope.

[0086] 2. The present application is based on the dynamic scaling of the partition range according to the crack risk type, which improves the monitoring coverage range of the high-risk area of the slope and compresses the coverage range of the low-risk area, thereby completely solving the problem of data fragmentation; and based on this, the scanning frequency of the synthetic aperture radar is dynamically adjusted, thereby improving the monitoring efficiency while reducing the cost. BRIEF DESCRIPTION OF DRAWINGS

[0087] Figure 1 The method flowchart of the present application. DETAILED DESCRIPTION

[0088] The present application provides a slope monitoring and early warning method based on oblique photography and radar technology, referring to Figure 1 A slope monitoring and early warning method based on oblique photography and radar technology, comprising:

[0089] S1: obtaining three-dimensional point cloud data of the slope by a multi-line laser radar; obtaining surface texture data of the slope by an oblique photography unmanned aerial vehicle; the surface texture data includes the position, size and quantity of the cracks on the slope surface;

[0090] S2: constructing a risk identification model, the model comprising: a registration and fusion module, an image recognition module and a comparison module;

[0091] Inputting the three-dimensional point cloud data and the surface texture data into the registration and fusion module for registration and fusion, to obtain three-dimensional image data of the slope;

[0092] Inputting the three-dimensional image data of the slope into the image recognition module for crack identification, to obtain a crack image, the crack image including the crack position, size and quantity;

[0093] The crack image is input into the comparison module, and the risk type is determined based on the preset threshold value and a graded alarm is issued;

[0094] S3: Perform zonal scanning and detection of the slope using synthetic aperture radar (SAR). Based on the crack image in S2, the zoning is optimized to obtain the optimized detection area. The scanning frequency of the SAR is dynamically adjusted based on the risk type of the optimized detection area.

[0095] Among them, the optimized detection area is obtained as follows:

[0096] ;

[0097] in, Indicates the optimized detection area; Represents the position coordinates of the pixel points in the optimized detection area; Represents cracks in the crack image The center position coordinates of Represents any crack in the crack image The center position coordinates of represents the Euclidean distance; Represents the crack in the corresponding crack image Type of risk The risk scaling factor of Represents any crack in the corresponding crack image Type of risk The risk scaling factor, represents the risk scaling factor.

[0098] Furthermore, in step S1, the three-dimensional point cloud data of the slope is obtained by multi-line laser radar as follows:

[0099] Use multi-line laser radar to set horizontal angular resolution, vertical angular resolution and point cloud density; obtain 3D point cloud data of the slope through multi-station scanning and stitching , expressed as:

[0100] ;

[0101] in, Indicates the first points Axis coordinates, Indicates the first points Axis coordinates, Indicates the first points axis coordinates;

[0102] The surface texture data of the slope is obtained by using an oblique photography drone in step S1, specifically:

[0103] The oblique photography drone flies crosswise at a preset inclination angle to generate orthophotos. The crack texture of the orthophotos is enhanced using the HSV algorithm to obtain the surface texture data of the slope.

[0104] Furthermore, the three-dimensional image data of the slope is obtained in step S2, specifically:

[0105] Spatial registration, expressed as:

[0106] ;

[0107] in, represents the rotation matrix, represents the translation vector, Represents surface texture data;

[0108] By minimizing the distance error between corresponding points , solve the rotation matrix and translation vector to geometrically align the surface texture data and the 3D point cloud data;

[0109] The registered surface texture data and 3D point cloud data are fused to obtain the 3D image data of the slope.

[0110] Furthermore, in step S2, the three-dimensional image data of the slope is input into the image recognition module for crack recognition to obtain a crack image, specifically:

[0111] Step S21: preprocessing and data enhancement of the three-dimensional image data of the slope to obtain an enhanced two-dimensional grayscale image;

[0112] Step S22: performing multi-scale feature extraction, attention enhancement, and decoding on the enhanced two-dimensional grayscale image to obtain a crack mask image, i.e., a crack image; in this process, obtaining a count label of the crack mask image;

[0113] Step S23: performing a morphological closing operation on the crack mask image to optimize the crack boundary, and performing noise filtering based on a threshold of the connected domain area and the image size, to finally obtain a crack mask image after noise filtering;

[0114] Step S24: obtaining the position, size and number of cracks based on the crack mask image after noise filtering.

[0115] Furthermore, step S21 is specifically as follows:

[0116] Preprocess the input slope 3D image data and expand the 3D image data into a 2D grayscale image , to preserve the brightness difference between the crack and the image background, expressed as;

[0117] ;

[0118] in, are the red channel value, green channel value and blue channel value of the three-dimensional image data respectively, All are preset weight parameters;

[0119] pass Algorithm to enhance two-dimensional grayscale images The local contrast of the image is enhanced to obtain the enhanced two-dimensional grayscale image. , expressed as:

[0120] ;

[0121] in, Represents the contrast-constrained adaptive histogram equalization function.

[0122] Furthermore, step S22 is specifically as follows:

[0123] Enhanced 2D grayscale image Perform multi-scale feature extraction to obtain multi-scale features , expressed as: ;

[0124] in, represents a lightweight convolutional layer used to extract the local texture of the crack; represents the dilated spatial pyramid pooling operation, which is used to capture the features of both fine and wide cracks;

[0125] Perform attention enhancement on multi-scale features to obtain enhanced multi-scale features , expressed as:

[0126] ;

[0127] in, Represents channel attention, which is used to suppress image interference;

[0128] Generate crack mask image through decoder , expressed as:

[0129] ;

[0130] in, Describes the decoder, represents the Sigmoid activation function,

[0131] Crack mask image Separate, obtain individual crack images and mark them, expressed as:

[0132] ;

[0133] in, Represents the connected domain marking function, which is used to distinguish adhesion or cross cracks. Represents the count label of a single crack image, Indicates the preset threshold.

[0134] Furthermore, step S23 is specifically as follows:

[0135] Crack mask image Morphological filtering is used to obtain the optimized crack mask image, which is expressed as: ;

[0136] in, Represents a closing operation; represents the dilation function, which is used to expand the boundary of the crack image; represents the erosion function, which is used to shrink the boundary of the crack image; represents the optimized crack mask image;

[0137] Filter the optimized crack mask image to obtain the crack mask image after noise filtering , expressed as:

[0138] ;

[0139] in, represents the indicator function, represents the connected domain area of ​​the input optimized crack mask image, represents the height of the input optimized crack mask image, represents the width of the input optimized crack mask image, is the preset noise figure.

[0140] Furthermore, step S24 is specifically as follows:

[0141] Get the crack mask image after noise filtering Location coordinates , expressed as:

[0142] ;

[0143] in, express The The crack coordinates of the pixel points, Represents the crack mask image after noise filtering Total number of pixels; Represents the connected domain area of ​​the input optimized crack mask image;

[0144] The crack mask image after noise filtering Perform size calculations to obtain the length of the crack , average width and maximum width , expressed as:

[0145] ;

[0146] ;

[0147] ;

[0148] in, Indicates the skeletonization of a single crack. A sequence of waypoints; Indicates the skeletonization of a single crack. A sequence of waypoints; represents the Euclidean distance; represents the ranging function;

[0149] The crack mask image after noise filtering Perform quantitative statistics to obtain the number of cracks , expressed as:

[0150] ;

[0151] in, Represents the maximum value function.

[0152] Furthermore, the process of determining whether to issue a risk warning based on a preset threshold is as follows:

[0153] Get risk factor , expressed as:

[0154] ;

[0155] in, 、 and are all preset weight parameters. Indicates the preset length of the crack, Indicates the preset maximum width of the crack. Indicates the maximum number of crack presets;

[0156] Based on risk factor Get risk type To conduct graded alarm, specifically:

[0157] ;

[0158] wherein 0, 1 and 2 are labels corresponding to respective risk types.

[0159] Further, the risk scaling factor in S3 corresponds to the risk type of the region, specifically:

[0160] When the risk type is a high-risk region, the risk scaling factor > 1;

[0161] When the risk type is a medium-risk region, the risk scaling factor = 1;

[0162] When the risk type is a low-risk region, the risk scaling factor < 1.

[0163] The above only describes the preferred embodiments of the present application and is not intended to limit the present application. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present application shall be included in the protection scope of the present application.

Claims

1. A slope monitoring and early warning method based on oblique photography and radar technology, characterized in that: include: S1: Acquire 3D point cloud data of the slope using multi-line laser radar; acquire surface texture data of the slope using oblique photography drones; Surface texture data include the location, size, and number of cracks on the slope surface; S2: Build a risk identification model, which includes: registration and fusion module, image recognition module and comparison module; Input the 3D point cloud data and surface texture data into the registration and fusion module for registration and fusion to obtain the 3D image data of the slope; Inputting the three-dimensional image data of the slope into the image recognition module to perform crack recognition and obtain a crack image, which includes the crack location, size and number; The crack image is input into the comparison module, and the risk type is determined based on the preset threshold value and a graded alarm is issued; S3: Perform zonal scanning and detection of the slope using synthetic aperture radar (SAR). Based on the crack image in S2, the zoning is optimized to obtain the optimized detection area. The scanning frequency of the SAR is dynamically adjusted based on the risk type of the optimized detection area. Among them, the optimized detection area is obtained as follows: ； in, Indicates the optimized detection area; Represents the position coordinates of the pixel points in the optimized detection area; Represents cracks in the crack image The center position coordinates of Represents any crack in the crack image The center position coordinates of represents the Euclidean distance; Represents the crack in the corresponding crack image Type of risk Risk scaling factor; Represents any crack in the corresponding crack image Type of risk The risk scaling factor, represents the risk scaling factor.

2. The slope monitoring and early warning method based on oblique photography and radar technology according to claim 1 is characterized in that: The three-dimensional point cloud data of the slope is obtained by multi-line laser radar in step S1 as follows: Use multi-line laser radar to set horizontal angular resolution, vertical angular resolution and point cloud density; obtain 3D point cloud data of the slope through multi-station scanning and stitching , expressed as: ； in, Indicates the first points Axis coordinates, Indicates the first points Axis coordinates, Indicates the first points axis coordinates; The surface texture data of the slope is obtained by using an oblique photography drone in step S1, specifically: The oblique photography drone flies crosswise at a preset inclination angle to generate orthophotos. The crack texture of the orthophotos is enhanced using the HSV algorithm to obtain the surface texture data of the slope.

3. The slope monitoring and early warning method based on oblique photography and radar technology according to claim 2 is characterized in that: The three-dimensional image data of the slope is obtained in step S2, specifically: Spatial registration, expressed as: ； in, represents the rotation matrix, represents the translation vector, Represents surface texture data; By minimizing the distance error between corresponding points , solve the rotation matrix and translation vector to geometrically align the surface texture data and the 3D point cloud data; The registered surface texture data and 3D point cloud data are fused to obtain the 3D image data of the slope.

4. The slope monitoring and early warning method based on oblique photography and radar technology according to claim 1 is characterized in that: In step S2, the three-dimensional image data of the slope is input into the image recognition module for crack recognition to obtain a crack image, specifically: Step S21: preprocessing and data enhancement of the three-dimensional image data of the slope to obtain an enhanced two-dimensional grayscale image; Step S22: performing multi-scale feature extraction, attention enhancement, and decoding on the enhanced two-dimensional grayscale image to obtain a crack mask image, i.e., a crack image; in this process, obtaining a count label of the crack mask image; Step S23: performing a morphological closing operation on the crack mask image to optimize the crack boundary, and performing noise filtering based on a threshold of the connected domain area and the image size, to finally obtain a crack mask image after noise filtering; Step S24: obtaining the position, size and number of cracks based on the crack mask image after noise filtering.

5. The slope monitoring and early warning method based on oblique photography and radar technology according to claim 4 is characterized in that: Step S21 is specifically as follows: Preprocess the input slope 3D image data and expand the 3D image data into a 2D grayscale image , to preserve the brightness difference between the crack and the image background, expressed as; ； in, are the red channel value, green channel value and blue channel value of the three-dimensional image data respectively, All are preset weight parameters; pass Algorithm to enhance two-dimensional grayscale images The local contrast of the image is enhanced to obtain the enhanced two-dimensional grayscale image. , expressed as: ； in, Represents the contrast-constrained adaptive histogram equalization function.

6. The slope monitoring and early warning method based on oblique photography and radar technology according to claim 5 is characterized in that: Step S22 is specifically as follows: Enhanced 2D grayscale image Perform multi-scale feature extraction to obtain multi-scale features , expressed as: ; in, represents a lightweight convolutional layer used to extract the local texture of the crack; represents the dilated spatial pyramid pooling operation, which is used to capture the features of both fine and wide cracks; Perform attention enhancement on multi-scale features to obtain enhanced multi-scale features , expressed as: ； in, Represents channel attention, which is used to suppress image interference; Generate crack mask image through decoder , expressed as: ； in, Describes the decoder, represents the Sigmoid activation function, Crack mask image Separate, obtain individual crack images and mark them, expressed as: ； in, Represents the connected domain marking function, which is used to distinguish adhesion or cross cracks. Represents the count label of a single crack image, Indicates the preset threshold.

7. The slope monitoring and early warning method based on oblique photography and radar technology according to claim 6 is characterized in that: Step S23 is specifically as follows: Crack mask image Morphological filtering is used to obtain the optimized crack mask image, which is expressed as: ; in, Represents a closing operation; represents the dilation function, which is used to expand the boundary of the crack image; represents the erosion function, which is used to shrink the boundary of the crack image; represents the optimized crack mask image; Filter the optimized crack mask image to obtain the crack mask image after noise filtering , expressed as: ； in, represents the indicator function, represents the connected domain area of ​​the input optimized crack mask image, represents the height of the input optimized crack mask image, represents the width of the input optimized crack mask image, is the preset noise figure.

8. The slope monitoring and early warning method based on oblique photography and radar technology according to claim 7 is characterized in that: Step S24 is specifically as follows: Get the crack mask image after noise filtering Location coordinates , expressed as: ； in, express The The crack coordinates of the pixel points, Represents the crack mask image after noise filtering Total number of pixels; Represents the connected domain area of ​​the input optimized crack mask image; The crack mask image after noise filtering Perform size calculations to obtain the length of the crack , average width and maximum width , expressed as: ； ； ； in, Indicates the skeletonization of a single crack. A sequence of waypoints; Indicates the skeletonization of a single crack. A sequence of waypoints; represents the Euclidean distance; represents the ranging function; The crack mask image after noise filtering Perform quantitative statistics to obtain the number of cracks , expressed as: ； in, Represents the maximum value function.

9. The slope monitoring and early warning method based on oblique photography and radar technology according to claim 8 is characterized in that: The specific process of determining whether to issue a risk alert based on the preset threshold is as follows: Get risk factor , expressed as: ； in, 、 and are all preset weight parameters. Indicates the preset length of the crack, Indicates the preset maximum width of the crack. Indicates the maximum number of crack presets; Based on risk factor Get risk type To conduct graded alarm, specifically: ； Among them, 0, 1, and 2 are the labels corresponding to each risk type.

10. The slope monitoring and early warning method based on oblique photography and radar technology according to claim 9, characterized in that: The risk scaling factors in S3 correspond to the risk types of the regions, specifically: When the risk type is high risk area, the risk scaling factor is >1; When the risk type is medium risk area, the risk scaling factor = 1; When the risk type is low risk area, the risk scaling factor is <1.

Images