A slope surface displacement patrol monitoring system and method based on a packet plane target point and multi-graph robust fusion

Abstract

This invention discloses a slope surface displacement inspection and monitoring system and method based on grouped planar target points and robust fusion of multiple images. The system includes: a target point layout module, a planar coordinate modeling module, an image acquisition and processing module, a target point identification module, a homography solution module, a matrix optimization module, a displacement inversion module, a fusion module, an early warning module, and a data management module. This invention uses uniquely coded target points, which significantly improves the reliability of automatic identification and cross-time-series matching of slope monitoring points. By grouping slope target points according to local coplanar features, this invention avoids errors caused by establishing a unified planar model for complex slopes as a whole, thus improving the accuracy of geometric modeling.

Description

Technical Field

[0001] This invention belongs to the field of slope safety monitoring technology, specifically relating to a slope surface displacement inspection and monitoring system and method based on grouped planar target points and robust fusion of multiple maps. Background Technology

[0002] Slopes are widely used in mining, highway and railway construction, water conservancy projects, building foundation pits, and natural mountain management. Affected by factors such as rainfall infiltration, groundwater activity, temperature changes, weathering and erosion, blasting vibrations, and excavation unloading, the surface and shallow structure of slopes undergo varying degrees of deformation, which can lead to landslides and collapses in severe cases. Therefore, long-term, continuous, low-cost, and highly reliable monitoring of slope displacement is of significant engineering importance.

[0003] Existing slope displacement monitoring methods mainly include contact monitoring methods, measuring instrument monitoring methods, and non-contact visual monitoring methods. Contact monitoring methods, such as displacement gauges, crack gauges, and inclinometers, offer high accuracy but are complex to deploy, have high maintenance costs, and limited coverage. Methods such as total stations and global navigation satellite systems can achieve high-precision measurements, but the equipment costs are high and the inspection efficiency is limited. While 3D laser scanning, synthetic aperture radar interferometry, and conventional photogrammetry can achieve large-area coverage, they often suffer from high costs, complex processing, or insufficient real-time performance in complex engineering sites.

[0004] In recent years, machine vision methods have attracted attention due to their low cost and ease of automated deployment. However, existing visual displacement monitoring technologies still have the following shortcomings in slope inspection applications: 1. The natural texture of slopes is greatly affected by lighting, weathering, occlusion, and seasonal changes, making them unreliable as stable feature points; 2. Images acquired by fixed cameras and drones have significant scale variations, pose variations, and perspective distortion, leading to unstable measurement results; 3. Slopes are usually not ideal single planes, and establishing a unified projection model for the entire slope surface will introduce significant geometric errors; 4. Single-image results are easily affected by blurring, occlusion, recognition errors, and local anomalies, resulting in a high false alarm rate; 5. Existing technologies lack a robust multi-image fusion and group alarm mechanism suitable for engineering inspection scenarios.

[0005] Therefore, there is an urgent need to provide a displacement inspection and monitoring system and method suitable for slope engineering sites, so as to realize the automatic identification, stable matching, local geometric inversion and reliable fusion of multiple monitoring target points, thereby improving the accuracy, robustness and early warning reliability of slope surface displacement monitoring. Summary of the Invention

[0006] This invention aims to address the shortcomings of existing technologies and provides the following solutions: A slope surface displacement inspection and monitoring system based on grouped planar target points and robust fusion of multiple images includes: a target point layout module, a planar coordinate modeling module, an image acquisition and processing module, a target point identification module, a homography solving module, a matrix optimization module, a displacement inversion module, a fusion module, an early warning module, and a data management module. The target placement module is used to place multiple marked target points on the slope surface; The planar coordinate modeling module performs planar modeling based on the marked target point to obtain the initial local planar coordinates of the marked target point; The image acquisition and processing module is used to acquire slope inspection images and perform validity screening on the slope inspection images to obtain multiple valid images; The target identification module is used to identify the code of the marked target in the slope inspection image and extract the corresponding current image coordinates; The homography solving module is used to construct multiple candidate homography matrices for each target group and each valid image based on the initial local plane coordinates and the current image coordinates; The matrix optimization module selects the optimal homography matrix from the candidate homography matrices based on the weighted projection error of all visible target points. The displacement inversion module is used to map image coordinates back to local plane coordinates using the inverse transformation of the optimal homography matrix, and to calculate the displacement results of each of the marked target points; The fusion module is used to robustly fuse the displacement results of the same marked target point in multiple valid images to obtain the final displacement of this inspection. The early warning module is used to determine whether to trigger an alarm based on the final displacement of the patrol and according to the displacement threshold and the number of target points exceeding the limit. The data management module is used to store the obtained target information, image information, displacement results, and alarm records.

[0007] Preferably, the workflow of the target deployment module includes: Before setting up the target points, a three-dimensional measurement of the natural surface of the slope is carried out to form a shape map of the slope surface, and a tight-fitting envelope surface that matches the natural surface of the slope is established. Based on the close-fitting envelope module, a suitable area for deploying marked target points is selected to obtain the target point deployment area; Based on local flatness, structural stability, constructability, image visibility, occlusion, and inspection view conditions, the target placement area is further screened to obtain the final target placement area. Multiple marked target points are deployed in the final target deployment area.

[0008] Preferably, the planar coordinate modeling module includes: a target point grouping unit, an initial measurement unit, and a planar modeling unit; The target point grouping unit is used to divide the marked target points into multiple target point groups according to the local geometric features of the slope. The initial measurement unit is used to obtain the initial three-dimensional coordinates of each of the marked target points; The planar modeling unit is used to fit a local plane based on the initial three-dimensional coordinates of each target group and establish a corresponding local two-dimensional coordinate system to obtain the initial local plane coordinates of each marked target.

[0009] Preferably, the image acquisition and processing module includes: an image acquisition unit and an image filtering unit; The image acquisition unit is used to acquire multiple images of the slope inspection. The image filtering unit is used to filter the slope inspection images based on image clarity and the number of target points to obtain multiple valid images.

[0010] Preferably, the workflow of the target identification module includes: The marked target points in the effective image are encoded and identified, and the corresponding numbers are parsed. Let the first In the target group, the first The marked target point at the ... The image coordinates in Zhang's effective image are: , in, Indicates the first k The first photo In the target group, point i of x Axis coordinates Indicates the first k The first photo In the target group, point i of y Axis coordinates; The homogeneous form is then expressed as: ; If the coordinates of the four corner points of the marked target are identified as follows: , , and Then, the geometric center of the four corner points is taken as the target image center to obtain the corresponding current image coordinates.

[0011] Preferably, the workflow of the homography solving module includes: For each of the target groups and each of the valid images, the first The target group in the first The set of target points in a valid image, denoted by the index set: , in, Indicates the first The target group in the first k In the photo One identifiable target point; The homography matrix is ​​obtained based on the initial local plane coordinates and the current image coordinates: , in, hmn This represents the elements in the homography matrix. m =1,2,3 n =1,2,3; Select multiple items of size from the index set. candidate subset : ; For the candidate subset any corresponding point The homogeneous linear equation is established using the direct linear transformation method: , , in, This represents the local horizontal coordinates of the target point on the slope plane in the initial state during target installation. This represents the local vertical coordinates of the target point on the slope plane in the initial state during target installation. Indicates matrix transpose; The candidate subset Stack the equations of all corresponding points in the matrix to form a coefficient matrix. Then we have: ; Through the Perform singular value decomposition and select the right singular vector corresponding to the minimum singular value as... The candidate homography matrix is ​​then reconstructed. .

[0012] Preferably, the workflow of the matrix optimization module includes: For candidate homography matrix The corresponding target group will be in the first Reproject all visible target points in the effective image; For any visible target point The predicted homogeneous coordinates of the image are: , in, Represents the homogeneous coordinates of the target point under the baseline condition; The predicted image coordinates are obtained after normalization: , in, Indicates the first Group 1 i The x-coordinate of the reprojection of the point's original coordinates onto the current image. Indicates the first Group 1 i The ordinate of the reprojection of the point's original coordinates onto the current image; The reprojection error is obtained based on the predicted image coordinates: ; The weighted error evaluation function for the candidate homography matrix is ​​obtained based on the reprojection error: , in, Indicates the target weight; The optimal homography matrix is ​​selected from all the candidate homography matrices by choosing the one with the smallest value of the weighted error evaluation function. , in, Indicates the first The target group in the first The set of candidate subsets in Zhang's valid images.

[0013] Preferably, the workflow of the displacement inversion module includes: After obtaining the optimal homography matrix Then, the target image coordinates are back-projected onto the local plane using the inverse matrix. For the first... In the target group, the first There are 10 target markers: ; After normalization, the current local plane coordinates are obtained: , in, This represents the x-coordinate of the projection of the current point onto the reference plane. The ordinate represents the projection of the current point onto the reference plane; Based on the current local plane coordinates, the target point is calculated at the [missing information]. Local planar displacement components under Zhang's effective image: , , , Based on the local planar displacement components, calculate the corresponding displacement modulus: .

[0014] Preferably, the workflow of the fusion module includes: For the In the target group, the first Mark target points to obtain a set of effective displacement modulus samples: , in, This indicates the number of valid observations of the marked target point during this patrol. Robust fusion is performed on the effective displacement modulus sample set to obtain the final displacement of the marked target point in this inspection: , in, This represents the median of the displacement sample set. This represents the minimum value in the set of displacement samples.

[0015] This invention also provides a slope surface displacement inspection and monitoring method based on grouped planar target points and robust fusion of multiple maps. The method is applied to the above-mentioned system and includes the following steps: S1. Set up multiple target points on the slope surface; S2. Perform planar modeling based on the marked target points to obtain the initial local planar coordinates of the marked target points; S3. Acquire slope inspection images and perform validity screening on the slope inspection images to obtain multiple valid images; S4. Identify the code of the marked target point in the slope inspection image and extract the corresponding current image coordinates; S5. For each target group and each effective image, construct multiple candidate homography matrices based on the initial local plane coordinates and the current image coordinates; S6. Based on the weighted projection error of all visible target points, select the optimal homography matrix from the candidate homography matrices; S7. The image coordinates are mapped back to local plane coordinates using the inverse transformation of the optimal homography matrix, and the displacement results of each of the marked target points are calculated; S8. Robustly fuse the displacement results of the same marked target point in multiple valid images to obtain the final displacement of this inspection; S9. Based on the final displacement of the inspection, and according to the displacement threshold and the number of target points exceeding the limit, determine whether to trigger an alarm; S10. Store the obtained target information, image information, displacement results, and alarm records.

[0016] Compared with the prior art, the beneficial effects of the present invention are as follows: (1) The present invention uses a marker target with a unique code, which can significantly improve the reliability of automatic identification and cross-time matching of slope monitoring points; (2) This invention avoids the errors caused by establishing a unified planar model for the entire complex slope by grouping the target points of the slope according to the local coplanar features, thereby improving the accuracy of geometric modeling; (3) The present invention improves the robustness to misidentification, local occlusion and anomalies by constructing multiple candidate homography matrices and performing optimization based on weighted projection error; (4) This invention improves the stability and reliability of the inspection results by robustly fusing the displacement results of the same target point in multiple valid images; (5) This invention reduces the false alarm rate caused by occasional errors at a single point by adopting a mechanism that triggers an alarm based on the number of points exceeding the limit; (6) The system of the present invention is compatible with various image acquisition methods such as fixed cameras and drones, and is applicable to various slope engineering scenarios such as mines, highways, railways, water conservancy and foundation pits. Attached Figure Description

[0017] To more clearly illustrate the technical solution of the present invention, the drawings used in the embodiments are briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0018] Figure 1 This is a schematic diagram of the system structure according to an embodiment of the present invention. Detailed Implementation

[0019] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0020] To make the above-mentioned objects, features and advantages of the present invention more apparent and understandable, the present invention will be further described in detail below with reference to the accompanying drawings and specific embodiments.

[0021] Example 1 In this embodiment, as Figure 1 As shown, a slope surface displacement inspection and monitoring system based on grouped planar target points and robust fusion of multiple images includes: a target point layout module, a planar coordinate modeling module, an image acquisition and processing module, a target point identification module, a homography solution module, a matrix optimization module, a displacement inversion module, a fusion module, an early warning module, and a data management module.

[0022] The target placement module is used to place multiple marked target points on the slope surface.

[0023] The workflow of the target placement module includes: before placing the marked targets, a three-dimensional measurement of the natural surface of the slope is performed to form a slope surface shape map, and a tight-fitting envelope surface adapted to the natural surface of the slope is established; based on the tight-fitting envelope module, a suitable area for placing marked targets is selected to obtain the target placement area; based on local flatness, structural stability, constructability, image visibility, occlusion, and inspection view conditions, the target placement area is further selected to obtain the final target placement area; multiple marked targets are placed in the final target placement area.

[0024] In this embodiment, before setting up the target points, it is preferable to first perform a three-dimensional measurement of the natural surface of the slope to form a slope surface shape map, and then establish a closely fitting envelope surface adapted to the natural surface of the slope. The three-dimensional measurement can be achieved using three-dimensional laser scanning, total station measurement, UAV photogrammetry, close-range photogrammetry, or a combination thereof. The slope surface shape map can be a point cloud model, a triangular mesh model, or a three-dimensional mesh model. The closely fitting envelope surface is a reference surface that fits and represents the natural surface of the slope; it can be a continuous surface or a piecewise continuous surface. The closely fitting envelope surface is used to select areas suitable for setting up target points. Preferably, the areas within the closely fitting envelope surface that coincide with the natural surface of the slope, and the areas whose shortest distance relative to the natural surface of the slope is no greater than 0.2 m, are determined as the target point placement areas.

[0025] If we take Represents the set of points on the natural surface of the slope, with If we represent the set of regions closely adhering to the envelope, then the target placement area can be represented as: , in, Represents a point on the envelope surface The shortest distance to the natural surface of the slope. The distance threshold is preferably 0.2 m. Based on this, the target placement area can be further screened by considering factors such as local flatness, structural stability, constructability, image visibility, occlusion, and inspection viewing angle, resulting in the final target placement area. This method ensures that the marked target points are rationally distributed along the natural shape of the slope, improving target installation stability, image recognition reliability, and displacement monitoring accuracy.

[0026] The planar coordinate modeling module performs planar modeling based on the marked target points to obtain the initial local planar coordinates of the marked target points.

[0027] The planar coordinate modeling module includes: a target point grouping unit, an initial measurement unit, and a planar modeling unit. The target point grouping unit is used to divide the marked target points into multiple target point groups based on the local geometric features of the slope; the initial measurement unit is used to obtain the initial three-dimensional coordinates of each marked target point; the planar modeling unit is used to fit a local plane based on the initial three-dimensional coordinates of each target point group and establish a corresponding local two-dimensional coordinate system to obtain the initial local planar coordinates of each marked target point.

[0028] In this embodiment, it is assumed that a total of [number] slope protection devices are distributed on the slope. The target group, the first Each target group contains One marked target point. In the target group, the first The initial three-dimensional coordinates of the target points are: , in, , For the first A group of target points, whose point set is represented as: ; To establish a local planar model, the centroid of the target point set is first calculated: , And construct a centered matrix: ; Perform singular value decomposition on the centered matrix, and use the right singular vector corresponding to the minimum singular value as the normal vector of the local plane: , Thus, the first Local plane equations corresponding to each target group: , ; Then, with the center of mass As the origin of the local planar coordinate system Choose two mutually orthogonal unit basis vectors in the local plane. and ,satisfy: , , , Therefore, the first In the target group, the first Projecting each marked target point into the local plane coordinate system yields the initial local plane coordinates: , , Therefore, the initial local plane coordinates of the marked target point are represented as: , Its homogeneous coordinates are represented as: ; The above steps can convert the original three-dimensional engineering coordinates into two-dimensional reference coordinates suitable for local planar homography modeling.

[0029] The image acquisition and processing module is used to acquire slope inspection images and filter the effectiveness of the slope inspection images to obtain multiple valid images.

[0030] The image acquisition and processing module includes an image acquisition unit and an image filtering unit. The image acquisition unit is used to acquire multiple slope inspection images; the image filtering unit is used to filter the slope inspection images based on image clarity and the number of target points identified, obtaining multiple valid images.

[0031] In this embodiment, during the inspection, the image acquisition module acquires multiple slope images. Assume that a total of [number] images are acquired during one inspection. Image number 1 Zhang Image is recorded as .

[0032] Each image is evaluated for quality. Preferably, Laplacian variance is used as the image sharpness index. Let the... The grayscale image of Zhang is Its Laplace response is denoted as: , The image sharpness index can then be defined as: , when When the image is determined to be blurry, it is removed. A preset resolution threshold is set. Simultaneously, the statistics for the [number]th [item] are [calculated / Total number of successfully identified target markers in the images .when When the image is deemed invalid, it is removed. This is a preset minimum number of recognition thresholds.

[0033] For any target group If the target group is in the first The number of target points identified in Zhang Youzheng's image is Then only if Only then is the homography matrix solution performed on the target group in the image.

[0034] The target identification module is used to identify the codes of marked target points in slope inspection images and extract the corresponding current image coordinates.

[0035] The workflow of the target identification module includes: The labeled target points in the valid image are encoded and identified, and the corresponding numbers are parsed.

[0036] Let the first In the target group, the first The marked target point at the ... The image coordinates in Zhang's effective image are: , in, Indicates the first k The first photo In the target group, point i of x Axis coordinates Indicates the first k The first photo In the target group, point i of y Axis coordinates. Then the homogeneous form is: ; If the coordinates of the four corner points of the marked target are identified as follows: , , and Then, the geometric centers of the four corner points are taken as the target image centers, and the corresponding current image coordinates are obtained, i.e.: ; In other implementations, the center of the border, the center of the circular target, the center of the cross, or the subpixel fitting center can also be used as the image coordinates.

[0037] The homography solution module is used to construct multiple candidate homography matrices for each target group and each valid image based on the initial local plane coordinates and the current image coordinates.

[0038] In this embodiment, the workflow of the homography solving module includes: For each target group and each valid image, the first The target group in the first The set of target points in a valid image, denoted by the index set: , in, Indicates the first The target group in the first k In the photo The first identifiable target point. Due to the first... Each marked target point in a target point group lies in the same local plane or approximately in the same local plane, and its initial local plane coordinates satisfy a projective transformation relationship with the image coordinates: , in, It is a non-zero scaling factor. For the first The target group in the first The homography matrix in a valid image is represented as: , in, hmn This represents the elements in the homography matrix. m =1,2,3 n =1,2,3. Since the homography matrix has 8 independent degrees of freedom, at least 4 pairs of non-collinear corresponding points are required to solve it. To improve robustness, this embodiment does not directly solve it based on all identification points at once, but instead selects multiple points of size from the index set. candidate subset : ; Preferred selection For each candidate subset The candidate homography matrix is ​​solved based on the initial local plane coordinates and image coordinates. .

[0039] Specifically, for candidate subsets any corresponding point The homogeneous linear equation is established using the direct linear transformation method: , , in, This represents the local horizontal coordinates of the target point on the slope plane in the initial state during target installation. This represents the local vertical coordinates of the target point on the slope plane in the initial state during target installation. This represents the matrix transpose. It transforms the candidate subset... Stack the equations of all corresponding points in the matrix to form a coefficient matrix. Then we have: ; Through the Perform singular value decomposition and select the right singular vector corresponding to the minimum singular value as... And reconstruct the candidate homography matrix. .

[0040] The matrix optimization module selects the optimal homography matrix from the candidate homography matrices based on the weighted projection error of all visible target points.

[0041] In this embodiment, the workflow of the matrix optimization module includes: For candidate homography matrix The corresponding target group will be in the first Reproject all visible target points in the valid image. For any visible target point... The predicted homogeneous coordinates of the image are: , in, These represent the homogeneous coordinates of the target point under the baseline condition. After normalization, the predicted image coordinates are obtained: , in, Indicates the first Group 1 i The x-coordinate of the reprojection of the point's original coordinates onto the current image. Indicates the first Group 1 i The ordinate of the reprojection of the point's original coordinates onto the current image. The reprojection error is obtained based on the predicted image coordinates. , Right now: ; To reflect the differences in importance, stability, or recognition reliability of different labeled targets, weights are assigned to each labeled target. The weighted error evaluation function for the candidate homography matrix is ​​obtained based on the reprojection error: , in, This represents the weight of the labeled target points. Alternatively, it can be expressed as the normalized weighted root mean square error. ; Then, the matrix with the smallest weighted error evaluation function value is selected from all candidate homography matrices as the optimal homography matrix: , in, Indicates the first The target group in the first The set of candidate subsets in Zhang's valid images.

[0042] In this embodiment, weight The weights are set based on the target's location, stability experience, recognition confidence, local image sharpness, and imaging scale. Preferably, the dynamic weights can be constructed as follows: , in, Indicates prior weights, Indicates the confidence level of identification. This represents a local image quality index. α , β , γ This represents the non-negative adjustment coefficient.

[0043] The displacement inversion module is used to map image coordinates back to local plane coordinates using the inverse transformation of the optimal homography matrix, and to calculate the displacement results of each marked target point.

[0044] In this embodiment, the workflow of the displacement inversion module includes: To obtain the optimal homography matrix Then, the target image coordinates are back-projected onto the local plane using the inverse matrix. For the first... In the target group, the first There are 10 target markers: ; After normalization, the current local plane coordinates are obtained: , in, This represents the x-coordinate of the projection of the current point onto the reference plane. This represents the ordinate of the projection of the current point onto the reference plane. The target point is calculated at the [missing coordinate] position based on the current local plane coordinates. Local planar displacement components under Zhang's effective image: , , , Calculate the corresponding displacement modulus based on the local planar displacement components: .

[0045] When it is necessary to output three-dimensional displacement in the engineering coordinate system, the local planar displacement can be mapped back to three-dimensional space, that is: , in, and is the unit basis vector of the corresponding local planar coordinate system.

[0046] The fusion module is used to robustly fuse the displacement results of the same marked target point in multiple valid images to obtain the final displacement of this inspection.

[0047] In this embodiment, the workflow of the fusion module includes: During a single inspection, the same marked target point may appear in multiple valid images, resulting in multiple displacement observations. For the first... In the target group, the first Mark target points to obtain a set of effective displacement modulus samples: , in, This indicates the number of valid observations of the marked target point during this inspection. To reduce the impact of outliers, this embodiment performs robust fusion on the effective displacement modulus sample set to obtain the final displacement of the marked target point during this inspection: , in, This represents the median of the displacement sample set. This represents the minimum value in the displacement sample set. At that time, take directly .when In this case, the average of the two results can be taken, or the above-mentioned fusion rule can still be used. In other embodiments, robust statistical methods such as weighted average, truncated mean, median and lower quartile average, and M-estimation can also be used, but this embodiment preferably uses the median and minimum average method.

[0048] The early warning module is used to determine whether to trigger an alarm based on the final displacement of the patrol, the displacement threshold, and the number of target points exceeding the limit.

[0049] In this embodiment, the workflow of the early warning module includes: Let the first In the target group, the first The displacement alarm threshold for each marker target point is: .when When this happens, the target marker is deemed to have exceeded the limit. Define an exceedance indicator variable: ; The total number of over-limit marked target points found during this inspection is as follows: ; when When this occurs, a system alarm is triggered, among which... This is the preset minimum alarm point threshold.

[0050] In one alternative implementation, group alarm rules can also be set independently for each target group. For the first... The number of out-of-limit targets in each target group is: ; when When this occurs, a group alarm is triggered for the area corresponding to that target group.

[0051] Furthermore, when historical inspection data is available, the displacement rate can also be calculated: ; when When this occurs, it can be determined as a rate over-limit alarm.

[0052] The data management module is used to store the obtained target information, image information, displacement results, and alarm records.

[0053] Example 2 In this embodiment, a slope surface displacement inspection and monitoring method based on grouped planar target points and robust fusion of multiple maps includes the following steps: S1. Set up multiple target points on the slope surface.

[0054] S2. Perform planar modeling based on the marked target points to obtain the initial local planar coordinates of the marked target points.

[0055] S3. Obtain slope inspection images and filter the images for validity to obtain multiple valid images.

[0056] S4. Identify the codes of the marked target points in the slope inspection images and extract the corresponding current image coordinates.

[0057] S5. For each target group and each valid image, construct multiple candidate homography matrices based on the initial local plane coordinates and the current image coordinates.

[0058] S6. Based on the weighted projection error of all visible target points, select the optimal homography matrix from the candidate homography matrices. S7. Use the inverse transformation of the optimal homography matrix to map the image coordinates back to the local plane coordinates, and calculate the displacement results of each marked target point.

[0059] S8. Robustly fuse the displacement results of the same marked target point in multiple valid images to obtain the final displacement of this inspection.

[0060] S9. Based on the final displacement of the patrol, determine whether to trigger an alarm according to the displacement threshold and the number of target points exceeding the limit.

[0061] S10. Store the obtained target information, image information, displacement results, and alarm records.

[0062] Example 3 In this embodiment, a specific implementation method is given.

[0063] A certain rock slope was divided into 4 local target point groups, namely .

[0064] Each target group has 12 marked target points, i.e. , .

[0065] The initial three-dimensional coordinates of each marked target point were measured using a total station, and a local plane was fitted to each group to establish a local plane coordinate system.

[0066] During a patrol, a total of 9 slope images were acquired. After image filtering, 6 valid images were retained. For the second target group, 10 marked target points were identified in one of the valid images. .

[0067] Multiple 4-point candidate subsets are constructed from these 10 marked target points, and the candidate homography matrix and weighted root mean square error are calculated for each subset. If we obtain: , , ; The candidate homography matrix corresponding to the minimum error of 1.15 is then selected as the optimal homography matrix.

[0068] Let the image coordinates of the 5th marked target in the second target group be: ; Its current local plane coordinates are obtained by inverse transformation of the optimal homography matrix: ; Its initial local plane coordinates are: ; The displacement components are as follows: , ; The corresponding displacement modulus is: ; If the displacement samples of the marked target point in the 6 valid images are: ; The median is: ; The minimum value is: ; Therefore, the final displacement of the marked target point during this inspection is: ; Possible options are: If the system's single-point displacement alarm threshold is set to 3.0 mm, and the system's minimum alarm point threshold is 3, then when the following conditions are met during this inspection... An alarm is triggered when there are at least 3 marked target points.

[0069] This invention is not limited to the above-described embodiments, and may also have the following variations: 1. The encoding format for marking target points can be replaced according to engineering needs; 2. The image acquisition device can be any device capable of acquiring slope images; 3. The candidate homography matrix can be generated by exhaustive combination, random sampling, or RANSAC method; 4. The weighting method can be fixed weight, empirical weight, or dynamic quality weight; 5. Robust fusion methods can be the average of the median and the minimum, or other robust statistics; 6. Alarm rules may include one or more combinations of displacement threshold, rate threshold, number of consecutive over-limits threshold, and group over-limit ratio threshold.

[0070] The embodiments described above are merely preferred embodiments of the present invention and are not intended to limit the scope of the present invention. Various modifications and improvements made by those skilled in the art to the technical solutions of the present invention without departing from the spirit of the present invention should fall within the protection scope defined by the claims of the present invention.

Claims

1. A slope surface displacement inspection and monitoring system based on grouped planar target points and robust fusion of multiple maps, characterized in that, include: The system includes a target placement module, a plane coordinate modeling module, an image acquisition and processing module, a target identification module, a homography solution module, a matrix optimization module, a displacement inversion module, a fusion module, an early warning module, and a data management module. The target placement module is used to place multiple marked target points on the slope surface; The planar coordinate modeling module performs planar modeling based on the marked target point to obtain the initial local planar coordinates of the marked target point; The image acquisition and processing module is used to acquire slope inspection images and perform validity screening on the slope inspection images to obtain multiple valid images; The target identification module is used to identify the code of the marked target in the slope inspection image and extract the corresponding current image coordinates; The homography solving module is used to construct multiple candidate homography matrices for each target group and each valid image based on the initial local plane coordinates and the current image coordinates; The matrix optimization module selects the optimal homography matrix from the candidate homography matrices based on the weighted projection error of all visible target points. The displacement inversion module is used to map image coordinates back to local plane coordinates using the inverse transformation of the optimal homography matrix, and to calculate the displacement results of each of the marked target points; The fusion module is used to robustly fuse the displacement results of the same marked target point in multiple valid images to obtain the final displacement of this inspection. The early warning module is used to determine whether to trigger an alarm based on the final displacement of the patrol and according to the displacement threshold and the number of target points exceeding the limit. The data management module is used to store the obtained target information, image information, displacement results, and alarm records.

2. The slope surface displacement inspection and monitoring system based on grouped planar target points and robust fusion of multiple maps according to claim 1, characterized in that, The workflow of the target deployment module includes: Before setting up the target points, a three-dimensional measurement of the natural surface of the slope is carried out to form a shape map of the slope surface, and a tight-fitting envelope surface that matches the natural surface of the slope is established. Based on the close-fitting envelope module, a suitable area for deploying marked target points is selected to obtain the target point deployment area; Based on local flatness, structural stability, constructability, image visibility, occlusion, and inspection view conditions, the target placement area is further screened to obtain the final target placement area. Multiple marked target points are deployed in the final target deployment area.

3. The slope surface displacement inspection and monitoring system based on grouped planar target points and robust fusion of multiple maps according to claim 1, characterized in that, The planar coordinate modeling module includes: a target point grouping unit, an initial measurement unit, and a planar modeling unit; The target point grouping unit is used to divide the marked target points into multiple target point groups according to the local geometric features of the slope. The initial measurement unit is used to obtain the initial three-dimensional coordinates of each of the marked target points; The planar modeling unit is used to fit a local plane based on the initial three-dimensional coordinates of each target group and establish a corresponding local two-dimensional coordinate system to obtain the initial local plane coordinates of each marked target.

4. The slope surface displacement inspection and monitoring system based on grouped planar target points and robust fusion of multiple maps according to claim 1, characterized in that, The image acquisition and processing module includes: an image acquisition unit and an image filtering unit; The image acquisition unit is used to acquire multiple images of the slope inspection. The image filtering unit is used to filter the slope inspection images based on image clarity and the number of target points to obtain multiple valid images.

5. The slope surface displacement inspection and monitoring system based on grouped planar target points and robust fusion of multiple maps according to claim 1, characterized in that, The workflow of the target identification module includes: The marked target points in the effective image are encoded and identified, and the corresponding numbers are parsed. Let the first In the target group, the first The marked target point at the ... The image coordinates in Zhang's effective image are: ， in, Indicates the first k The first photo In the target group, point i of x Axis coordinates Indicates the first k The first photo In the target group, point i of y Axis coordinates; The homogeneous form is then expressed as: ； If the coordinates of the four corner points of the marked target are identified as follows: , , and Then, the geometric center of the four corner points is taken as the target image center to obtain the corresponding current image coordinates.

6. The slope surface displacement inspection and monitoring system based on grouped planar target points and robust fusion of multiple maps according to claim 5, characterized in that, The workflow of the homography solving module includes: For each of the target groups and each of the valid images, the first The target group in the first The set of target points in a valid image, denoted by the index set: ， in, Indicates the first The target group in the first k In the photo One identifiable target point; The homography matrix is ​​obtained based on the initial local plane coordinates and the current image coordinates: ， in, hmn This represents the elements in the homography matrix. m =1,2,3 n =1,2,3; Select multiple items of size from the index set. candidate subset : ； For the candidate subset any corresponding point The homogeneous linear equation is established using the direct linear transformation method: ， ， in, This represents the local horizontal coordinates of the target point on the slope plane in the initial state during target installation. This represents the local vertical coordinates of the target point on the slope plane in the initial state during target installation. Indicates matrix transpose; The candidate subset Stack the equations of all corresponding points in the matrix to form a coefficient matrix. Then we have: ； Through the Perform singular value decomposition and select the right singular vector corresponding to the minimum singular value as... The candidate homography matrix is ​​then reconstructed. .

7. The slope surface displacement inspection and monitoring system based on grouped planar target points and robust fusion of multiple maps according to claim 6, characterized in that, The workflow of the matrix optimization module includes: For candidate homography matrix The corresponding target group will be in the first Reproject all visible target points in the effective image; For any visible target point The predicted homogeneous coordinates of the image are: ， in, Represents the homogeneous coordinates of the target point under the baseline condition; The predicted image coordinates are obtained after normalization: ， in, Indicates the first Group 1 i The x-coordinate of the reprojection of the point's original coordinates onto the current image. Indicates the first Group 1 i The ordinate of the reprojection of the point's original coordinates onto the current image; The reprojection error is obtained based on the predicted image coordinates: ； The weighted error evaluation function for the candidate homography matrix is ​​obtained based on the reprojection error: ， in, Indicates the target weight; The optimal homography matrix is ​​selected from all the candidate homography matrices by choosing the one with the smallest value of the weighted error evaluation function. ， in, Indicates the first The target group in the first The set of candidate subsets in Zhang's valid images.

8. The slope surface displacement inspection and monitoring system based on grouped planar target points and robust fusion of multiple maps according to claim 7, characterized in that, The workflow of the displacement inversion module includes: After obtaining the optimal homography matrix Then, the target image coordinates are back-projected onto the local plane using the inverse matrix. For the first... In the target group, the first There are 10 target markers: ； After normalization, the current local plane coordinates are obtained: ， in, This represents the x-coordinate of the projection of the current point onto the reference plane. The ordinate represents the projection of the current point onto the reference plane; Based on the current local plane coordinates, the target point is calculated at the [missing information]. Local planar displacement components under Zhang's effective image: ， ， ， Based on the local planar displacement components, calculate the corresponding displacement modulus: 。 9. The slope surface displacement inspection and monitoring system based on grouped planar target points and robust fusion of multiple maps according to claim 8, characterized in that, The workflow of the fusion module includes: For the In the target group, the first Mark target points to obtain a set of effective displacement modulus samples: ， in, This indicates the number of valid observations of the marked target point during this patrol. Robust fusion is performed on the effective displacement modulus sample set to obtain the final displacement of the marked target point in this inspection: ， in, This represents the median of the displacement sample set. This represents the minimum value in the set of displacement samples.

10. A method for slope surface displacement inspection and monitoring based on grouped planar target points and robust fusion of multiple maps, wherein the method is applied to the system described in any one of claims 1-9, characterized in that, Includes the following steps: S1. Set up multiple target points on the slope surface; S2. Perform planar modeling based on the marked target points to obtain the initial local planar coordinates of the marked target points; S3. Acquire slope inspection images and perform validity screening on the slope inspection images to obtain multiple valid images; S4. Identify the code of the marked target point in the slope inspection image and extract the corresponding current image coordinates; S5. For each target group and each effective image, construct multiple candidate homography matrices based on the initial local plane coordinates and the current image coordinates; S6. Based on the weighted projection error of all visible target points, select the optimal homography matrix from the candidate homography matrices; S7. The image coordinates are mapped back to local plane coordinates using the inverse transformation of the optimal homography matrix, and the displacement results of each of the marked target points are calculated; S8. Robustly fuse the displacement results of the same marked target point in multiple valid images to obtain the final displacement of this inspection; S9. Based on the final displacement of the inspection, and according to the displacement threshold and the number of target points exceeding the limit, determine whether to trigger an alarm; S10. Store the obtained target information, image information, displacement results, and alarm records.

Images