A slope displacement detection method based on an anti-occlusion and weathering-resistant passive three-dimensional target

Abstract

The application discloses a kind of based on anti-occlusion anti-weathering passive three-dimensional target's slope displacement detection method, it is related to slope displacement monitoring technical field, comprising: by monitoring equipment real-time shooting target initial image;Target initial image is executed pre-processing operation;Calculate the occlusion rate and determine whether the target image after pre-processing is effective, divide no-occlusion working condition and local occlusion working condition;No-occlusion adopts subpixel level target center positioning algorithm, local occlusion adopts coding fragment matching algorithm, obtain target center coordinates;Compare target center coordinates with initial three-dimensional reference coordinates, calculate three-dimensional displacement data;Three-dimensional displacement data are compared with preset displacement early warning threshold, if exceed, trigger sound-light early warning.The application provides a kind of slope displacement detection method and stereoscopic passive three-dimensional target structure, from structure and method dual dimension solves the problem that conventional target and detection method is easy to fail in slope field environment, monitoring dimension is single, poor anti-occlusion.

Description

Technical Field

[0001] This invention relates to the field of slope displacement monitoring technology, specifically a slope displacement detection method based on a passive three-dimensional target that is resistant to obstruction and weathering. Background Technology

[0002] With the continuous advancement of geological disaster prevention and control, slope displacement monitoring, as a core means of early warning for geological disasters such as landslides and roadbed collapses, is widely used in scenarios such as highway and railway slopes, mine slopes, reservoir slopes, and mountain building slopes. Target displacement detection technology has become one of the mainstream technologies for slope displacement monitoring due to its ease of operation and strong real-time performance.

[0003] Conventional detection schemes often employ planar targets paired with pixel-level detection algorithms to monitor and warn of slope displacement. However, slope monitoring areas such as mountainous and mining areas are often located in harsh outdoor environments characterized by dense vegetation, abundant rainfall, large temperature differences, and frequent sandstorms. The application of conventional planar targets and their associated detection methods in these scenarios presents several challenges: 1. Planar targets are often made of paper or ordinary spray-painted materials, which are easily faded by rain and weathered by wind and sand, leading to blurring and failure of the markings within a short period. 2. Planar targets only possess two-dimensional planar positioning features, enabling the acquisition of only horizontal two-dimensional displacement data. This single monitoring dimension fails to monitor vertical and tilt displacements of slopes, potentially resulting in incomplete disaster warnings. 3. Conventional detection algorithms provide pixel-level positioning, which is insufficient to meet the precise early warning requirements for minute slope displacements. 4. Natural vegetation growth on outdoor slopes can cause partial occlusion of the targets. Conventional detection algorithms lack specific recognition compensation mechanisms; even slight occlusion can cause target positioning failure, hindering the continuous acquisition of displacement data and impacting the accuracy of early warnings for slope geological disasters.

[0004] To address these issues, existing technologies primarily optimize target materials or detection algorithms to improve slope target monitoring performance. However, they have not developed a comprehensive solution to resist obstruction and weathering in harsh outdoor environments, nor have they addressed the three-dimensional monitoring problem from the perspective of target structure. Consequently, they fail to meet the monitoring requirements for high-precision, multi-dimensional, and high-stability slope displacement.

[0005] Based on this, a slope displacement detection method based on a passive three-dimensional target that is resistant to shading and weathering is now provided, which can eliminate the drawbacks of existing technical solutions. Summary of the Invention

[0006] The purpose of this invention is to provide a slope displacement detection method based on an anti-shading and anti-weathering passive three-dimensional target, so as to solve the problems in the background art.

[0007] To achieve the above objectives, the present invention provides the following technical solution: A slope displacement detection method based on a passive three-dimensional target that is resistant to shading and weathering includes the following steps: Step S1: Fix passive three-dimensional targets and monitoring equipment at each monitoring point on the slope. Take initial images of the targets in real time and mark the initial three-dimensional reference coordinates through the monitoring equipment. The targets are integral injection molded structures of weather-resistant engineering plastics, and irregular concave and convex coded dot matrix is ​​arranged on the outer surface. Step S2: Perform preprocessing operations on the initial image of the target to enhance the contour features of the target's concave and convex coding dot matrix and eliminate interference from outdoor light and shadow, dust accumulation, and background. The preprocessing includes at least noise reduction filtering, grayscale normalization, and edge feature enhancement operations. Step S3: Detect the occlusion ratio of the target region in the preprocessed target image using an image recognition algorithm, calculate the occlusion rate, determine whether the preprocessed target image is effective based on the occlusion rate, and divide the corresponding effective preprocessed target images into unoccluded and partially occluded conditions based on the occlusion determination threshold. Step S4: Under unobstructed conditions, a sub-pixel-level target center localization algorithm is adopted. Gray-level interpolation is performed on the edge pixels of the target area through quadratic interpolation, and the edge of the target area is fitted and optimized through least squares to solve the target center coordinates and obtain the real-time three-dimensional spatial coordinates of the target. Under partial occlusion conditions, a coding fragment matching algorithm is adopted to extract the fragment features of the concave and convex coding dot matrix of the unoccluded area in the preprocessed target image. The fragment features are matched with the preset target standard coding feature library for similarity. After successful matching, the target center coordinates are inferred from the fragment features to realize the target localization under occlusion conditions. Step S5: Compare the real-time acquired three-dimensional spatial coordinates of the target with the initial three-dimensional reference coordinates, and calculate the three-dimensional displacement data of the slope horizontal X / Y axis displacement, vertical Z axis settlement displacement, and target tilt angle displacement respectively, so as to achieve the effect of three-dimensional displacement monitoring. Step S6: Upload the three-dimensional displacement data to the background monitoring terminal and compare it with the preset displacement warning threshold. If the three-dimensional displacement data exceeds the preset displacement warning threshold, trigger an audible and visual warning. Simultaneously store the entire process data of three-dimensional displacement data acquisition, preprocessing, calculation, and comparison to form a displacement ledger. The displacement ledger is used for slope disaster source tracing analysis.

[0008] Furthermore, the target includes a target body, an embossed coding dot matrix, and a mounting base. The target body is designed as a three-dimensional frustum structure with a temperature resistance range of -40℃ to 80℃, a height of 80mm, and a bottom diameter of 100mm. The outer surface of the target body is injection molded with an embossed coding dot matrix, the height of the raised dots of the embossed coding dot matrix is ​​2mm, the bottom diameter is 3mm, and the dot matrix distribution is a unique random code. The bottom of the target body is fixedly connected to the mounting base, and the lower surface of the mounting base is provided with several expansion bolt holes for fixing to the slope monitoring point with stainless steel expansion bolts.

[0009] Furthermore, the specific process of the preprocessing operation in step S2 includes: The initial image of the target is smoothed using a Gaussian filtering algorithm to remove random noise caused by wind and sand and light fluctuations in the wild, while retaining the edge details of the concave and convex coded dot matrix. Perform an overall grayscale unification operation on the target image after noise reduction and filtering to eliminate the overall brightness difference caused by morning and evening, shade, and cloudy days, and standardize the pixel grayscale features of the concave and convex coding dot matrix. Gradient operators are used to extract the edge contours of the convex and concave coding dot matrix, and the difference in grayscale between the convex dots of the convex and concave coding dot matrix and the base grayscale of the target image is amplified to distinguish the target from the slope background.

[0010] Further, in step S3, the occlusion ratio of the target region in the preprocessed target image is detected by an image recognition algorithm, and the occlusion rate is calculated. The specific process includes: In the target image after noise reduction filtering, grayscale normalization and edge feature enhancement preprocessing, the region of interest completely covered by the target is delineated. Based on the target body parameters and monitoring equipment parameters, the total theoretical pixel area corresponding to the complete concave and convex coding dot matrix of the target in the region of interest is calculated. The pixels in the region of interest are classified and determined by foreground segmentation, distinguishing between valid pixels of the convex-concave coding dot matrix and occluded invalid pixels, and filtering out all occluded invalid pixels. The distinction formula is expressed as follows: ,in, For the binary label value of the pixel in the region of interest, Indicates the effective pixel count. Indicates that an invalid pixel is being occluded. Coordinates of the preprocessed target image Location pixel grayscale values, grayscale value range is uniformly normalized , This is the segmentation threshold; Traverse all occluded and invalid pixels within the region of interest, and calculate the total area of ​​these occluded and invalid pixels. Then, calculate the real-time occlusion rate based on the sum of the total area of ​​the invalid pixels and the theoretical total pixel area. The formula for calculating the occlusion rate is as follows: ,in, For the real-time occlusion rate of the target, To mask the total area of ​​invalid pixels, This represents the theoretical total pixel area.

[0011] Further, in step S3, the validity of the preprocessed target image is determined based on the occlusion rate. The corresponding valid preprocessed target images are divided into unoccluded and partially occluded conditions based on the occlusion determination threshold. The specific process includes: the background monitoring terminal has a preset occlusion determination threshold, which is set to 35%. When the occlusion rate is greater than the occlusion determination threshold, the preprocessed target image is determined to be invalid, the current image is discarded and re-acquired. If multiple consecutive image acquisitions are all invalid images, the slope displacement calculation is paused and the occlusion alarm information is uploaded to the background monitoring terminal. When the occlusion rate is not greater than the occlusion determination threshold, the preprocessed target image is determined to be valid. If the occlusion rate is zero, it is determined to be an unoccluded condition. If the occlusion rate is between zero and the occlusion determination threshold, it is determined to be a partially occluded condition.

[0012] Furthermore, the calculation process of the sub-pixel-level target center localization algorithm in step S4 specifically includes: Extract the edge pixel coordinates corresponding to all the concave and convex coding points on the target surface under unobstructed conditions, and collect the discrete pixel information of the target contour. Perform secondary grayscale interpolation along the horizontal and vertical dimensions of the edge pixels of the target region to fill the grayscale information between discrete pixels and supplement sub-pixel level details; The least squares method is used to fit the curve of the continuous target edge contour after interpolation, correcting the positioning deviation caused by discrete pixels, and obtaining a smooth and continuous target closed contour. The target geometric center is solved based on the fitted target closed contour, and the sub-pixel level plane center coordinates are output. By combining the target body parameters and monitoring equipment parameters, the sub-pixel-level planar center coordinates are converted into real-time target three-dimensional spatial coordinates.

[0013] Furthermore, the calculation process of the coded fragment matching algorithm in step S4 specifically includes: The concave-convex encoded dot matrix of the unoccluded area in the pre-processed target image is located, and the convex coordinates, convex spacing, arrangement order and convex size are extracted to form the fragment feature vector; Retrieve the pre-stored target standard coding feature library from the background monitoring terminal. The target standard coding feature library contains complete concave and convex coding dot matrix segment feature data of all targets at the monitoring point. The fragment feature vector is compared with the complete segmented feature data in the target standard encoded feature library one by one to calculate the similarity. A similarity matching threshold is set. When the similarity is higher than the similarity matching threshold, the fragment feature is determined to belong to the corresponding target. After a successful match, the complete concave-convex coding dot matrix distribution of the target is completed in reverse based on the target body parameters and arrangement order. The edge fitting optimization is performed on the contour of the completed concave-convex coding dot matrix using the least squares method. Combined with the monitoring equipment parameters and target body parameters, the target center coordinates are deduced in reverse.

[0014] Furthermore, the calculation process for the three-dimensional displacement data in step S5 specifically includes: Read the initial three-dimensional reference coordinates pre-stored in step S1, including the initial X-axis reference, Y-axis reference, Z-axis reference, and initial tilt angle reference; The difference between the target's three-dimensional spatial coordinates and the initial three-dimensional reference coordinates is calculated to obtain the horizontal lateral displacement of the X-axis, the horizontal longitudinal displacement of the Y-axis, and the vertical settlement displacement of the Z-axis, respectively. Compare the real-time target's three-dimensional spatial deflection angle with the initial tilt angle reference to calculate the target's tilt angle; By integrating triaxial displacement data with target tilt angle data, complete three-dimensional displacement data of the slope is generated.

[0015] Furthermore, the background monitoring terminal and the monitoring equipment communicate with each other via a network, and the background monitoring terminal is used for: Receive real-time three-dimensional displacement data and visualize the displacement changes at each monitoring point on the slope; Store all process data to form a displacement ledger and support historical data retrieval and export; Storage algorithm, occlusion determination threshold, target standard encoding feature library and preset displacement warning threshold; Upon receiving the obstruction alarm information, it synchronously outputs an audible and visual control command to activate the on-site audible and visual alarm device. The analysis of landslide causes can be achieved by retrieving single-frame images, preprocessing parameters, and displacement records.

[0016] Furthermore, in step S1, the monitoring device adopts an ultra-black light high-definition imaging device with a resolution of not less than 4K, a minimum illumination of 0.001 lux, and a fixed image acquisition frame rate of 1 frame / second.

[0017] Compared with the prior art, the beneficial effects of the present invention are as follows: 1. This invention features a passive three-dimensional target that is resistant to obstruction and weathering. The target body is made of corrosion-resistant and anti-aging engineering plastic through one-piece injection molding, achieving passive operation and maintenance-free operation. The target has a three-dimensional structure with three-dimensional spatial positioning features, which can capture horizontal, vertical, and tilt displacement information in all dimensions. The surface of the target body is covered with an irregular concave-convex code dot matrix. The dot matrix is ​​a physically formed structure without any spray coating, which can avoid fading due to rainwater erosion and abrasion from wind and sand. Moreover, the concave-convex code has unique identification, which facilitates fragment matching under obstruction conditions. 2. This invention is equipped with a subpixel-level target center positioning algorithm. Combined with the high-definition image of the target acquired by the monitoring equipment, the subpixel-level positioning of the target center is achieved through image grayscale interpolation, edge fitting optimization and other methods, thereby improving the monitoring accuracy, identifying small displacement changes of the slope, and realizing early warning of geological disasters. 3. This invention is equipped with a coded fragment matching algorithm. For scenarios where vegetation partially obscures the target or the target is partially damaged, the algorithm extracts the concave and convex coded dot matrix fragment features of the unobscured area of ​​the target and matches them with a preset target standard coded feature library. This algorithm can accurately locate and calculate the displacement of the target when the obscuration rate does not exceed 35%, thus solving the problem of detection failure caused by dense vegetation on slopes. Attached Figure Description

[0018] Figure 1 This is a schematic diagram of the method steps of the present invention.

[0019] Figure 2 This is a schematic diagram of the target structure of the present invention.

[0020] Figure 3 This is a schematic diagram illustrating the relationship between the target, monitoring equipment, and back-end monitoring terminal of the present invention.

[0021] Figure label annotations: Target 100, Target body 110, Embossed coding dot matrix 120, Mounting base 130, Monitoring equipment 200, Back-end monitoring terminal 300. Detailed Implementation

[0022] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to the accompanying drawings and embodiments.

[0023] Example 1

[0024] In this embodiment, as Figure 1 As shown, this invention provides a slope displacement detection method based on an anti-shading and anti-weathering passive three-dimensional target, specifically including the following steps: Step S1: Fix passive three-dimensional targets and monitoring equipment 200 at each monitoring point on the slope. Take initial images of the targets in real time and mark the initial three-dimensional reference coordinates through the monitoring equipment 200. The target 100 is a weather-resistant engineering plastic one-piece injection molded structure with irregular concave and convex coding dot matrix on the outer surface. Step S2: Perform preprocessing operations on the initial target image to enhance the contour features of the target's concave and convex coding dot matrix and eliminate interference from outdoor lighting, dust accumulation, and background. The preprocessing includes at least noise reduction filtering, grayscale normalization, and edge feature enhancement operations. Step S3: Detect the occlusion ratio of the target region in the preprocessed target image using an image recognition algorithm, calculate the occlusion rate, determine whether the preprocessed target image is effective based on the occlusion rate, and divide the corresponding effective preprocessed target images into unoccluded and partially occluded conditions based on the occlusion determination threshold. Step S4: Under unobstructed conditions, a sub-pixel-level target center localization algorithm is adopted. Gray-level interpolation is performed on the edge pixels of the target area through quadratic interpolation, and the edge of the target area is fitted and optimized through least squares to solve the target center coordinates and obtain the real-time three-dimensional spatial coordinates of the target. Under partial occlusion conditions, a coding fragment matching algorithm is adopted to extract the fragment features of the concave and convex coding dot matrix of the unoccluded area in the preprocessed target image. The fragment features are matched with the preset target standard coding feature library for similarity. After successful matching, the target center coordinates are inferred from the fragment features to realize the target localization under occlusion conditions. Step S5: Compare the real-time acquired three-dimensional spatial coordinates of the target with the initial three-dimensional reference coordinates, and calculate the three-dimensional displacement data of the slope horizontal X / Y axis displacement, vertical Z axis settlement displacement, and target tilt angle displacement respectively, so as to achieve the effect of three-dimensional displacement monitoring. Step S6: Upload the three-dimensional displacement data to the background monitoring terminal 300 and compare it with the preset displacement warning threshold. If the three-dimensional displacement data exceeds the preset displacement warning threshold, an audible and visual warning will be triggered. The entire process of three-dimensional displacement data acquisition, preprocessing, calculation, and comparison will be stored synchronously to form a displacement ledger. The displacement ledger is used for slope disaster source tracing analysis.

[0025] Specifically, step S2, the preprocessing operation, includes: The initial target image is smoothed using a Gaussian filtering algorithm to remove random noise caused by wind and sand in the wild and light fluctuations, while preserving the edge details of the concave and convex coding dot matrix. Perform an overall grayscale unification operation on the target image after noise reduction and filtering to eliminate the overall brightness difference caused by morning and evening, shade, and cloudy days, and standardize the pixel grayscale features of the concave and convex coding dot matrix. Gradient operators are used to extract the edge contours of the convex and concave coding dot matrix, and the difference in grayscale between the convex dots of the convex and concave coding dot matrix and the base grayscale of the target image is amplified to distinguish the target from the slope background.

[0026] Specifically, in step S3, the occlusion ratio of the target region in the preprocessed target image is detected using an image recognition algorithm, and the occlusion rate is calculated, including: In the target image after noise reduction filtering, grayscale normalization, and edge feature enhancement preprocessing, the region of interest (ROI) completely covered by the target is delineated. Based on the target body parameters and monitoring equipment parameters, the theoretical total pixel area corresponding to the complete concave and convex coded dot matrix of the target within the ROI is calculated. The target body parameters are fixed factory-defined parameters, pre-stored in the background monitoring terminal 300, and do not change with the shooting environment. They can be used as the geometric reference for three-dimensional coordinate inverse projection and dot matrix completion, such as the overall physical dimensions of the target, the total number of concave and convex coded dot matrices, the diameter of a single convex dot, the center distance of the convex dot array, the target thickness, and the factory-defined complete three-dimensional geometric reference dimensions of the target. The monitoring equipment parameters are fixed parameters for camera calibration and on-site setup, used to convert two-dimensional image pixels into the real three-dimensional spatial coordinates of the slope, such as monitoring camera intrinsic parameters, camera extrinsic parameters, camera imaging resolution, lens focal length, shooting distance, image grayscale normalization range, and camera calibration distortion coefficient. The pixels in the region of interest are classified and determined by foreground segmentation, distinguishing between valid pixels of the convex-concave coding dot matrix and occluded invalid pixels, and filtering out all occluded invalid pixels. The distinction formula is expressed as follows: ,in, For the binary label value of the pixel in the region of interest, This represents the effective dot matrix pixels (convex and concave encoded convex areas, identifiable features). This indicates occluded invalid pixels (areas occluded by vegetation, dust, shadows, or debris, with no valid encoded features). Coordinates of the preprocessed target image Location pixel grayscale values, grayscale value range is uniformly normalized , The segmentation threshold is set to 30-80, and the range is determined based on statistical analysis of a large number of images with different lighting conditions and different obstructions in the field. It can be adjusted according to actual environmental requirements. Traverse all occluded and invalid pixels within the region of interest, calculate the total area of ​​occluded and invalid pixels, and then calculate the real-time occlusion rate based on the sum of the total area of ​​occluded and invalid pixels and the theoretical total pixel area. The formula for calculating the occlusion rate is as follows: ,in, For the real-time occlusion rate of the target, To mask the total area of ​​invalid pixels, The formula is the theoretical total pixel area. It can distinguish between the target's own concave and convex coding and external occlusions, making it easier to count the pixels in the occluded area and realize image validity and working condition classification. In step S3, the validity of the preprocessed target image is determined based on the occlusion rate. The corresponding valid preprocessed target images are divided into unoccluded and partially occluded conditions according to the occlusion determination threshold. This includes: the background monitoring terminal 300 has a preset occlusion determination threshold, which is set to 35%. The occlusion determination threshold is a critical value that has been experimentally verified to ensure the accuracy of the subsequent positioning algorithm. When the occlusion rate is greater than the occlusion determination threshold, the preprocessed target image is determined to be invalid, the current image is discarded and re-acquired. If multiple consecutive image acquisitions are all invalid images, the slope displacement calculation is paused and the occlusion alarm information is uploaded to the background monitoring terminal 300. When the occlusion rate is not greater than the occlusion determination threshold, the preprocessed target image is determined to be valid. If the occlusion rate is zero, it is determined to be an unoccluded condition. If the occlusion rate is between zero and the occlusion determination threshold, it is determined to be a partially occluded condition.

[0027] Specifically, the sub-pixel-level target center localization algorithm in step S4 can break through the limitation of pixel-level detection accuracy, achieving a monitoring accuracy of 0.01mm, which is far higher than the 0.1mm detection accuracy of conventional targets. This facilitates the accurate identification of minute displacement changes on slopes and enables early and accurate warnings of geological disasters. The calculation process of the sub-pixel-level target center localization algorithm specifically includes: Extract the edge pixel coordinates corresponding to all the concave and convex coding points on the target surface under unobstructed conditions. Collect discrete pixel information of the target outline. This represents the total number of discrete edge pixels extracted. Secondary grayscale interpolation is performed along the horizontal and vertical dimensions of the edge pixels of the target region to fill the grayscale information between discrete pixels and supplement sub-pixel level details. The formula for calculating secondary grayscale interpolation is as follows: ,in, integer pixels Grayscale value, This represents the grayscale value at the sub-pixel position after interpolation. and This is the sub-pixel offset. and It is a quadratic interpolation basis function; The least squares method is used to fit the interpolated, continuous target edge contour to correct the positioning deviation caused by discrete pixels, resulting in a smooth and continuous closed target contour. The implicit general equation of the ellipse is: The least squares fitting loss function is: , This represents the total number of sub-pixel contour points after interpolation and encryption. These are the sub-pixel contour coordinates after interpolation. , , , , , The fitting coefficients of the ellipse to be solved; Based on the fitted target closed contour, the target geometric center is solved, and the sub-pixel level planar center coordinates are output. (Center of the ellipse); Combined with the camera intrinsic parameter matrix of the monitoring equipment extrinsic rotation matrix With translation vector The physical dimensions of the target body are determined by using a pinhole camera inverse projection model to obtain sub-pixel-level planar center coordinates. Converted to real-time target 3D spatial coordinates ; The coded fragment matching algorithm can perform feature matching on coded fragments in unoccluded areas of the target, solving the target detection failure problem caused by dense vegetation on slopes and complex terrain. The specific calculation process of the coded fragment matching algorithm includes: After locating the convex-convex encoded dot matrix of the unoccluded area in the preprocessed target image, the convex coordinates, convex spacing, arrangement order, and convex size are extracted to form a fragment feature vector. ,in, To encode multidimensional fragment feature vectors locally after occlusion, the four-dimensional components correspond to the convex point coordinates, convex point spacing, arrangement order, and convex point size, respectively. This is the set of coordinates of convex points in the unobstructed area. Let be the set of Euclidean distances between any two convex points. The encoding value is the order of the convex dots. These are the normalized eigenvalues ​​of the physical dimensions of the convex points. For the unobstructed area, the first The coordinates of each convex point This represents the current number of unobstructed, valid bumps. The pre-stored target standard coding feature library is retrieved from the background monitoring terminal 300. This library contains complete concave-convex coding dot matrix segmented feature data for all targets at the monitoring points. The single-segment standard segmented feature vector is denoted as... subscript Representing the Standard segmented feature data of each target , , , The first The coordinates of the bumps, the spacing between the bumps, the arrangement order, and the size of the bumps for each target; The similarity is calculated by comparing the fragmented feature vectors with the complete segmented feature data in the target standard encoded feature library one by one. The calculation formula is as follows: ,in, The similarity between fragment feature vectors and standard segmented feature vectors. The inner product of the fragment feature vector and the standard segment feature vector is given. and These are the L2 norms of the fragment feature vector and the standard segmented feature vector, respectively. The first fragment feature vector dimensional components, The first piecewise feature vector of the standard segment Dimensional components, with a similarity matching threshold set. When the similarity is higher than the similarity matching threshold, Then determine the target corresponding to the fragment feature, i.e., the first... One target; After successful matching, the complete concave-convex coding dot matrix distribution of the target is completed in reverse based on the target body parameters and arrangement order. The completed concave-convex coding dot matrix is ​​represented as follows: , For the complete concave-convex coding dot matrix, the first The coordinates of each convex point To determine the total number of convex points in the complete target encoding, the edge fitting optimization is performed using the least squares method on the outline of the completed concave-convex encoding dot matrix after completion. The implicit equation of the ellipse is: The least squares loss function is Combined with the camera intrinsic parameter matrix of the monitoring equipment extrinsic rotation matrix With translation vector The physical dimensions of the target body are determined using a pinhole imaging inverse projection model. The coordinates of the target center can be obtained by reverse calculation. , among which, is, The subpixel center coordinates of the fitted image. For imaging scale factor, , , , , , The coefficients of the target contour ellipse fitting equation are used to solve for the sub-pixel center of the image. ; Specifically, the convex point coordinates are the coordinates of the convex points in the unoccluded area under partial occlusion conditions. The convex point coordinates are divided into two categories: image pixel coordinates and physical world coordinates. The image pixel coordinates are the two-dimensional pixel positions corresponding to the center of each convex point in the preprocessed target image. The physical world coordinates are the three-dimensional real space coordinates corresponding to each convex point on the target entity. The convex point spacing is the Euclidean distance between the centers of any two convex point codes, including the image pixel spacing and the entity physical spacing. It is calculated from the above image pixel coordinates and physical world coordinates respectively and is not affected by overall translation or scaling. The arrangement order is the convex point coding sequence of the convex points along the target clockwise or counterclockwise. It is uniquely determined by the polar angle of the convex point relative to the geometric center of the target. It is used to reverse the complete target point distribution by relying on the arrangement order when the convex point is lost due to partial occlusion. The convex point size is the normalized feature value of a single convex point code. It is used to filter out vegetation and stone noise points with large differences from the convex point pixel size and avoid invalid pixels from being mixed into the feature vector.

[0028] Specifically, the calculation process for the three-dimensional displacement data in step S5 includes: Read the initial three-dimensional reference coordinates pre-stored in step S1 Includes the initial X-axis reference Y-axis reference Z-axis reference Initial tilt angle reference ; Target three-dimensional spatial coordinates With the initial three-dimensional reference coordinates By performing interpolation calculations, the horizontal lateral displacement (X-axis), the horizontal longitudinal displacement (Y-axis), and the vertical settlement displacement (Z-axis) are obtained. The calculation formulas are as follows: , This represents the horizontal displacement along the X-axis. The horizontal longitudinal displacement along the Y-axis. The vertical settlement displacement along the Z-axis is compared with the real-time target's three-dimensional spatial deflection angle. Compared with the initial tilt angle reference Calculate the target tilt angle By integrating triaxial displacement data with target tilt angle data, complete three-dimensional slope displacement data is generated. .

[0029] In this embodiment, the preset displacement warning threshold in step S6 can be set as a preset multi-level displacement warning threshold group, which combines the slope displacement change rate and cumulative displacement amplitude to divide the multi-level warning levels. The formula for calculating the displacement change rate is: , The rate of change of displacement. This represents the displacement difference between adjacent acquisition cycles. The image acquisition time interval is used to match the corresponding early warning level based on the cumulative displacement amplitude range and displacement change rate range. Then, the real-time three-dimensional displacement data, the real-time target occlusion status, the matched early warning level, and the image acquisition timestamp are packaged and uploaded to the background monitoring terminal 300. The background monitoring terminal 300 performs differentiated actions according to the early warning level, such as corresponding level audio and visual push notifications, full-process data archiving and storage, and archiving of slope anomaly tracing records.

[0030] Example 2

[0031] The difference from Example 1 is that, as in Example 1, Figure 2 As shown, this invention provides a passive three-dimensional target that is resistant to obstruction and weathering, applied to the slope displacement detection method shown in Embodiment 1 above. It includes a target body 110, an embossed coding dot matrix 120, and a mounting base 130. The target body 110 is a three-dimensional frustum structure, integrally molded from engineering plastic (such as PA66+GF30). This material possesses corrosion resistance, aging resistance, high and low temperature resistance (-40℃~80℃), and UV resistance, making it suitable for harsh outdoor slope environments and solving the problem of conventional targets easily fading and failing. Its temperature resistance range is -40℃~80℃, its height is 80mm, and its bottom diameter is 100mm. The outer surface of the target body 110 is injection-molded with an embossed coding dot matrix 120. The embossed coding dot matrix 120 is a physical structure; even if the target surface has slight dust accumulation or dirt, it can still maintain clear feature markings. Without a sprayed coating, it can effectively resist rainwater erosion, wind and sand abrasion, and UV radiation, ensuring the stability of target identification and displacement detection. The embossed coding dot matrix 120... The protrusion height is 2mm, the bottom diameter is 3mm, and the dot matrix distribution is uniquely randomly coded. The bottom of the target body 110 is fixedly connected to the mounting base 130. Expansion bolt holes are drilled at the slope monitoring point, and the target 100 is fixed to the rock / concrete surface of the monitoring point by stainless steel expansion bolts. During installation, ensure that the target faces the imaging equipment and there is no large area of ​​initial obstruction. After installation, no debugging is required, achieving passive plug-and-play. The lower surface of the mounting base 130 has several expansion bolt holes for fixing to the slope monitoring point with stainless steel expansion bolts. The overall structure of the target 100 is passive and has no electronic components, which is to achieve full maintenance-free operation and a service life far longer than conventional targets, thereby reducing the operation and maintenance cost of slope monitoring. Moreover, the target 100 has a three-dimensional structure with three-dimensional spatial positioning characteristics. Combined with the detection algorithm, it can realize the synchronous monitoring of the horizontal, vertical, and tilt three-dimensional displacement of the slope landslide, making up for the deficiency of conventional planar targets that can only monitor two dimensions, and improving the accuracy and comprehensiveness of geological disaster early warning. Specifically, such as Figure 3As shown, the slope displacement detection method of this invention achieves slope displacement detection through the cooperation of a background monitoring terminal 300 and a monitoring device 200 via a target 100. The monitoring device 200 adopts an ultra-black light high-definition imaging device with a resolution of not less than 4K, a minimum illumination of 0.001 lux, and a fixed image acquisition frame rate of 1 frame / second. The background monitoring terminal 300 and the monitoring device 200 communicate with each other via a network. The background monitoring terminal 300 is used for: Receive real-time three-dimensional displacement data and visualize the displacement changes at each monitoring point on the slope; Store all process data to form a displacement ledger and support historical data retrieval and export; Storage algorithm, occlusion determination threshold, target standard encoding feature library and preset displacement warning threshold; Upon receiving the obstruction alarm information, it synchronously outputs an audible and visual control command to activate the on-site audible and visual alarm device. The analysis of landslide causes can be achieved by retrieving single-frame images, preprocessing parameters, and displacement records. It can also be used for: real-time transmission of image acquisition frequency, noise reduction filtering parameters, similarity matching threshold, and segmentation threshold to front-end monitoring devices 200; updating image recognition algorithm running parameters; uploading occlusion rate; and updating target standard coding feature library and early warning threshold, etc.

[0032] In summary, this invention provides a slope displacement detection method based on a passive three-dimensional target that is resistant to obstruction and weathering, and provides a three-dimensional passive three-dimensional target structure. It solves the problems of conventional targets and detection methods being prone to failure, having a single monitoring dimension, low accuracy, and poor resistance to obstruction in the field environment of slopes from the dual dimensions of target structure and detection method, and has good application prospects.

[0033] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the technical scope disclosed in this application should be included within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.

Claims

1. A slope displacement detection method based on an anti-shading and anti-weathering passive three-dimensional target, characterized in that, Specifically, the following steps are included: Step S1: Fix passive three-dimensional targets and monitoring equipment at each monitoring point on the slope. Take initial images of the targets in real time and mark the initial three-dimensional reference coordinates through the monitoring equipment. The targets are integral injection molded structures of weather-resistant engineering plastics, and irregular concave and convex coded dot matrix is ​​arranged on the outer surface. Step S2: Perform preprocessing operations on the initial image of the target to enhance the contour features of the target's concave and convex coding dot matrix and eliminate interference from outdoor light and shadow, dust accumulation, and background. The preprocessing includes at least noise reduction filtering, grayscale normalization, and edge feature enhancement operations. Step S3: Detect the occlusion ratio of the target region in the preprocessed target image using an image recognition algorithm, calculate the occlusion rate, determine whether the preprocessed target image is effective based on the occlusion rate, and divide the corresponding effective preprocessed target images into unoccluded and partially occluded conditions based on the occlusion determination threshold. Step S4: Under unobstructed conditions, a sub-pixel-level target center localization algorithm is adopted. Gray-level interpolation is performed on the edge pixels of the target area through quadratic interpolation, and the edge of the target area is fitted and optimized through least squares to solve the target center coordinates and obtain the real-time three-dimensional spatial coordinates of the target. Under partial occlusion conditions, a coding fragment matching algorithm is adopted to extract the fragment features of the concave and convex coding dot matrix of the unoccluded area in the preprocessed target image. The fragment features are matched with the preset target standard coding feature library for similarity. After successful matching, the target center coordinates are inferred from the fragment features to realize the target localization under occlusion conditions. Step S5: Compare the real-time acquired three-dimensional spatial coordinates of the target with the initial three-dimensional reference coordinates, and calculate the three-dimensional displacement data of the slope horizontal X / Y axis displacement, vertical Z axis settlement displacement, and target tilt angle displacement respectively, so as to achieve the effect of three-dimensional displacement monitoring. Step S6: Upload the three-dimensional displacement data to the background monitoring terminal and compare it with the preset displacement warning threshold. If the three-dimensional displacement data exceeds the preset displacement warning threshold, trigger an audible and visual warning. Simultaneously store the entire process data of three-dimensional displacement data acquisition, preprocessing, calculation, and comparison to form a displacement ledger. The displacement ledger is used for slope disaster source tracing analysis.

2. The slope displacement detection method based on an anti-shading and anti-weathering passive three-dimensional target according to claim 1, characterized in that, The target includes a target body, a raised and recessed coding dot matrix, and a mounting base. The target body is designed as a three-dimensional frustum structure with a temperature resistance range of -40℃ to 80℃, a height of 80mm, and a bottom diameter of 100mm. The outer surface of the target body is injection molded with a raised and recessed coding dot matrix, the height of the raised dots of the raised and recessed coding dot matrix is ​​2mm, the bottom diameter is 3mm, and the dot matrix distribution is a unique random code. The bottom of the target body is fixedly connected to the mounting base, and the lower surface of the mounting base is provided with several expansion bolt holes for fixing to the slope monitoring point with stainless steel expansion bolts.

3. The slope displacement detection method based on an anti-shading and anti-weathering passive three-dimensional target according to claim 1, characterized in that, The specific process of the preprocessing operation in step S2 includes: The initial image of the target is smoothed using a Gaussian filtering algorithm to remove random noise caused by wind and sand and light fluctuations in the wild, while retaining the edge details of the concave and convex coded dot matrix. Perform an overall grayscale unification operation on the target image after noise reduction and filtering to eliminate the overall brightness difference caused by morning and evening, shade, and cloudy days, and standardize the pixel grayscale features of the concave and convex coding dot matrix. Gradient operators are used to extract the edge contours of the convex and concave coding dot matrix, and the difference in grayscale between the convex dots of the convex and concave coding dot matrix and the base grayscale of the target image is amplified to distinguish the target from the slope background.

4. The slope displacement detection method based on an anti-shading and anti-weathering passive three-dimensional target according to claim 1, characterized in that, In step S3, the occlusion ratio of the target region in the preprocessed target image is detected by an image recognition algorithm, and the occlusion rate is calculated. The specific process includes: In the target image after noise reduction filtering, grayscale normalization and edge feature enhancement preprocessing, the region of interest completely covered by the target is delineated. Based on the target body parameters and monitoring equipment parameters, the total theoretical pixel area corresponding to the complete concave and convex coding dot matrix of the target in the region of interest is calculated. The pixels in the region of interest are classified and determined by foreground segmentation, distinguishing between valid pixels of the convex-concave coding dot matrix and occluded invalid pixels, and filtering out all occluded invalid pixels. The distinction formula is expressed as follows: ,in, For the binary label value of the pixel in the region of interest, Indicates the effective pixel count. Indicates that an invalid pixel is being occluded. Coordinates of the preprocessed target image Location pixel grayscale values, grayscale value range is uniformly normalized , The segmentation threshold; Traverse all occluded and invalid pixels within the region of interest, and calculate the total area of ​​these occluded and invalid pixels. Then, calculate the real-time occlusion rate based on the sum of the total area of ​​the invalid pixels and the theoretical total pixel area. The formula for calculating the occlusion rate is as follows: ,in, For the real-time occlusion rate of the target, To mask the total area of ​​invalid pixels, This represents the theoretical total pixel area.

5. The slope displacement detection method based on an anti-shading and anti-weathering passive three-dimensional target according to claim 1, characterized in that, In step S3, the validity of the preprocessed target image is determined based on the occlusion rate. The corresponding valid preprocessed target images are divided into unoccluded and partially occluded conditions according to the occlusion determination threshold. The specific process includes: the background monitoring terminal has a preset occlusion determination threshold, which is set to 35%. When the occlusion rate is greater than the occlusion determination threshold, the preprocessed target image is determined to be invalid, the current image is discarded and re-acquired. If multiple consecutive image acquisitions are all invalid images, the slope displacement calculation is paused and the occlusion alarm information is uploaded to the background monitoring terminal. When the occlusion rate is not greater than the occlusion determination threshold, the preprocessed target image is determined to be valid. If the occlusion rate is zero, it is determined to be an unoccluded condition. If the occlusion rate is between zero and the occlusion determination threshold, it is determined to be a partially occluded condition.

6. The slope displacement detection method based on an anti-shading and anti-weathering passive three-dimensional target according to claim 1, characterized in that, The calculation process of the sub-pixel level target center localization algorithm in step S4 specifically includes: Extract the edge pixel coordinates corresponding to all the concave and convex coding points on the target surface under unobstructed conditions, and collect the discrete pixel information of the target contour. Perform secondary grayscale interpolation along the horizontal and vertical dimensions of the edge pixels of the target region to fill the grayscale information between discrete pixels and supplement sub-pixel level details; The least squares method is used to fit the curve of the continuous target edge contour after interpolation, correcting the positioning deviation caused by discrete pixels, and obtaining a smooth and continuous target closed contour. The target geometric center is solved based on the fitted target closed contour, and the sub-pixel level plane center coordinates are output. By combining the target body parameters and monitoring equipment parameters, the sub-pixel-level planar center coordinates are converted into real-time target three-dimensional spatial coordinates.

7. The slope displacement detection method based on an anti-shading and anti-weathering passive three-dimensional target according to claim 1, characterized in that, The calculation process of the coded fragment matching algorithm in step S4 specifically includes: The concave-convex encoded dot matrix of the unoccluded area in the pre-processed target image is located, and the convex coordinates, convex spacing, arrangement order and convex size are extracted to form the fragment feature vector; Retrieve the pre-stored target standard coding feature library from the background monitoring terminal. The target standard coding feature library contains complete concave and convex coding dot matrix segment feature data of all targets at the monitoring point. The fragment feature vector is compared with the complete segmented feature data in the target standard encoded feature library one by one to calculate the similarity. A similarity matching threshold is set. When the similarity is higher than the similarity matching threshold, the fragment feature is determined to belong to the corresponding target. After a successful match, the complete concave-convex coding dot matrix distribution of the target is completed in reverse based on the target body parameters and arrangement order. The edge fitting optimization is performed on the contour of the completed concave-convex coding dot matrix using the least squares method. Combined with the monitoring equipment parameters and target body parameters, the target center coordinates are deduced in reverse.

8. The slope displacement detection method based on an anti-shading and anti-weathering passive three-dimensional target according to claim 1, characterized in that, The calculation process for the three-dimensional displacement data in step S5 specifically includes: Read the initial three-dimensional reference coordinates pre-stored in step S1, including the initial X-axis reference, Y-axis reference, Z-axis reference, and initial tilt angle reference; The difference between the target's three-dimensional spatial coordinates and the initial three-dimensional reference coordinates is calculated to obtain the horizontal lateral displacement of the X-axis, the horizontal longitudinal displacement of the Y-axis, and the vertical settlement displacement of the Z-axis, respectively. Compare the real-time target's three-dimensional spatial deflection angle with the initial tilt angle reference to calculate the target's tilt angle; By integrating triaxial displacement data with target tilt angle data, complete three-dimensional slope displacement data is generated.

9. The slope displacement detection method based on an anti-shading and anti-weathering passive three-dimensional target according to claim 1, characterized in that, The background monitoring terminal and the monitoring equipment communicate with each other via a network. The background monitoring terminal is used for: Receive real-time three-dimensional displacement data and visualize the displacement changes at each monitoring point on the slope; Store all process data to form a displacement ledger and support historical data retrieval and export; Storage algorithm, occlusion determination threshold, target standard coding feature library and preset displacement warning threshold; Upon receiving the obstruction alarm information, it simultaneously outputs an audible and visual control command to activate the on-site audible and visual alarm device. The analysis of landslide causes can be achieved by retrieving single-frame images, preprocessing parameters, and displacement records.

10. A slope displacement detection method based on an anti-shading and anti-weathering passive three-dimensional target according to claim 1, characterized in that, In step S1, the monitoring device is an ultra-black light high-definition imaging device with a resolution of not less than 4K, a minimum illumination of 0.001 lux, and a fixed image acquisition frame rate of 1 frame / second.

Images