Intelligent Inspection System for Construction Quality of Slope Anchor Drilling Platform Based on Machine Vision

Abstract

This application relates to the field of machine vision technology and discloses an intelligent inspection system for the construction quality of a slope anchor drilling platform based on machine vision. The method includes: an acquisition module acquiring original images with spatiotemporal markers from multiple angles; a correction module preprocessing to obtain standardized images; an extraction module constructing a three-dimensional model of the drilling platform; a monitoring module applying an improved adaptive Kalman filter algorithm to detect trajectory deviations; a calculation module evaluating drilling quality; and an analysis module comprehensively analyzing and generating quality levels and early warning information. This application constructs an intelligent inspection method for the construction quality of a slope anchor drilling platform based on machine vision, realizing real-time monitoring, dynamic analysis, and quality early warning of the construction process, overcoming the subjectivity and lag of traditional manual inspection, and improving the accuracy and efficiency of anchor drilling construction quality inspection.

Description

Technical Field

[0001] This application relates to the field of machine vision technology, and in particular to an intelligent inspection system for the construction quality of a slope anchor drilling platform based on machine vision. Background Technology

[0002] Slope anchor drilling technology is a commonly used reinforcement measure in slope engineering. It involves drilling anchor holes on the slope surface, installing anchor bolts, and grouting them in place to form an anchoring system that improves slope stability. Traditionally, the quality inspection of slope anchor drilling platforms relies mainly on manual visual inspection and simple tools, such as using a tape measure to measure drilling depth, using a level to check platform levelness, and testing anchoring force. With the development of computer vision and sensor technology, some projects have begun to introduce cameras to record the construction process and analyze the results to evaluate construction quality. In recent years, some research has attempted to apply technologies such as laser scanning and 3D reconstruction to slope engineering monitoring, but most studies focus on monitoring overall slope deformation; research on intelligent inspection of drilling platform construction quality remains relatively lacking.

[0003] Existing technologies have significant shortcomings: First, manual inspection methods are highly subjective, have low measurement accuracy, and cannot obtain quality information about the borehole interior and the anchor installation process. Second, existing machine vision applications are mostly for single-parameter detection or static image processing, lacking the ability to dynamically monitor and comprehensively analyze the entire construction process. Third, there is a gap between data acquisition and quality assessment, making real-time feedback and early warning difficult. Fourth, the correlation analysis between different quality parameters is lacking, making it impossible to identify the root causes of quality problems from a systemic perspective. Fifth, the inspection results lack objective quantitative standards and a systematic evaluation system, making it difficult to provide a scientific basis for construction decisions. These shortcomings lead to frequent quality problems in slope anchor projects, increasing project risks and maintenance costs. Summary of the Invention

[0004] This application provides a machine vision-based intelligent inspection system for the construction quality of slope anchor drilling platforms. It is used to construct an intelligent inspection method for the construction quality of slope anchor drilling platforms based on machine vision, realize real-time monitoring, dynamic analysis and quality early warning of the construction process, overcome the subjectivity and lag of traditional manual inspection, and improve the inspection accuracy and efficiency of anchor drilling construction quality.

[0005] This application provides a machine vision-based intelligent inspection system for the construction quality of slope anchor drilling platforms. The machine vision-based intelligent inspection system for the construction quality of slope anchor drilling platforms includes:

[0006] The acquisition module is used to acquire multi-angle images of the construction area of ​​the slope anchor drilling platform using a high-resolution industrial camera, and obtain a raw image dataset with spatiotemporal tags.

[0007] The correction module is used to perform image preprocessing and geometric correction on the original image dataset to obtain a standardized drilling platform image;

[0008] The extraction module is used to extract the geometric feature parameters of the drilling platform based on the standardized drilling platform image to obtain a three-dimensional spatial model of the drilling platform.

[0009] The monitoring module is used to apply an improved adaptive Kalman filter algorithm to the three-dimensional spatial model of the drilling platform to monitor the drilling trajectory in real time and obtain drilling trajectory deviation data.

[0010] The calculation module is used to calculate key parameters of borehole quality based on the borehole trajectory deviation data, and obtain borehole quality evaluation indicators.

[0011] The analysis module is used to comprehensively analyze the drilling quality evaluation indicators and the monitoring data of the anchor bolt installation process to obtain the construction quality level and early warning information of the slope anchor bolt drilling platform.

[0012] The technical solution provided in this application acquires images from multiple angles using a high-resolution industrial camera to obtain a raw image dataset with spatiotemporal markers. This effectively solves the problems of incomplete and discontinuous data from traditional single-point observations, providing rich raw data resources for subsequent comprehensive quality analysis. Image preprocessing and geometric correction are performed on the raw image dataset to obtain standardized drilling platform images. This overcomes interference factors such as unstable image quality and distorted perspectives in complex field environments, significantly improving the accuracy of subsequent analysis and processing. Based on the standardized drilling platform images, geometric feature parameters of the drilling platform are extracted to obtain a three-dimensional spatial model of the drilling platform. This transforms two-dimensional image information into three-dimensional spatial geometric information, enabling precise quantification of key parameters such as drilling platform size, drilling position, and angle. An improved adaptive Kalman filter algorithm is applied to the three-dimensional spatial model of the drilling platform to perform drilling trajectory simulation. Real-time monitoring yields borehole trajectory deviation data. The algorithm is optimized for the unique geological variations characteristic of slope engineering. Through real-time adjustment of the state noise covariance matrix, it significantly improves the robustness and accuracy of trajectory monitoring under complex geological conditions, enabling timely capture of minute deviations during the drilling process. Based on the borehole trajectory deviation data, key borehole quality parameters are calculated, resulting in borehole quality evaluation indicators. This achieves objective quantification of multi-dimensional quality characteristics such as borehole straightness, depth, diameter, and wall roughness, avoiding the subjectivity and uncertainty of traditional manual measurements. A comprehensive analysis of the borehole quality evaluation indicators and anchor bolt installation process monitoring data yields the construction quality level and early warning information for the slope anchor bolt drilling platform. Through multi-source data fusion and a comprehensive evaluation model, a systematic and standardized quality assessment system is established, achieving a leap from single-indicator detection to comprehensive quality evaluation. Of particular note is the improved adaptive Kalman filter algorithm introduced in the borehole trajectory monitoring stage, which fully considers the specific characteristics of slope engineering applications. This algorithm automatically adjusts the filter parameters by sensing changes in geological conditions, effectively solving the problem of large trajectory prediction deviations in complex geological areas using traditional Kalman filtering. In the comprehensive quality evaluation stage, the weight allocation mechanism combining entropy weighting and analytic hierarchy process (AHP) is perfectly suited to the multidimensional characteristics of slope anchor bolt engineering quality evaluation, overcoming the limitations of arbitrary subjective weighting and the inability of fixed weights to reflect actual changes. These domain-specific optimizations and combined applications of artificial intelligence algorithms give this solution significant advantages in the field of slope anchor bolt borehole quality detection, improving both the accuracy and comprehensiveness of detection, and enabling intelligent real-time monitoring and early warning, providing strong technical support for improving the construction quality and safety of slope engineering. Attached Figure Description

[0013] To more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings used in the description of the embodiments will be briefly introduced below. Obviously, the drawings described below are some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0014] Figure 1 This is a schematic diagram of one embodiment of the intelligent inspection system for construction quality of slope anchor drilling platform based on machine vision in this application. Detailed Implementation

[0015] This application provides a machine vision-based intelligent inspection system for the construction quality of slope anchor drilling platforms. The terms "first," "second," "third," "fourth," etc. (if present) in the specification, claims, and accompanying drawings of this application are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that such data can be interchanged where appropriate so that the embodiments described herein can be implemented in a sequence other than that illustrated or described herein. Furthermore, the terms "comprising" or "having" and any variations thereof are intended to cover a non-exclusive inclusion; for example, a process, method, system, product, or device that comprises a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or units not explicitly listed or inherent to such processes, methods, products, or devices.

[0016] For ease of understanding, the specific process of the embodiments of this application is described below. Please refer to [link / reference]. Figure 1 One embodiment of the intelligent inspection system for construction quality of slope anchor drilling platform based on machine vision in this application includes:

[0017] The acquisition module 101 is used to acquire multi-angle images of the construction area of ​​the slope anchor drilling platform using a high-resolution industrial camera, and obtain a raw image dataset with spatiotemporal markers.

[0018] The correction module 102 is used to perform image preprocessing and geometric correction on the original image dataset to obtain a standardized drilling platform image;

[0019] Extraction module 103 is used to extract the geometric feature parameters of the drilling platform based on the standardized drilling platform image to obtain a three-dimensional spatial model of the drilling platform.

[0020] Monitoring module 104 is used to apply an improved adaptive Kalman filter algorithm to the three-dimensional spatial model of the drilling platform to monitor the drilling trajectory in real time and obtain drilling trajectory deviation data.

[0021] The calculation module 105 is used to calculate key parameters of borehole quality based on the borehole trajectory deviation data, and obtain borehole quality evaluation indicators.

[0022] The analysis module 106 is used to comprehensively analyze the drilling quality evaluation indicators and the monitoring data of the anchor installation process to obtain the construction quality level and early warning information of the slope anchor drilling platform.

[0023] Specifically, the acquisition module 101 achieves multi-angle image acquisition of the construction area of ​​the slope anchor drilling platform. This is achieved by establishing a spatial coordinate system around the construction area using at least four positioning markers, and deploying at least three 12-megapixel global shutter CCD sensor industrial cameras to form a multi-angle coverage network. During acquisition, the industrial cameras continuously capture images at a frequency of 5 frames per second, simultaneously recording image timestamps and spatial location information. For critical construction stages such as the start and completion of drilling, the acquisition frequency automatically increases to 15 frames per second to capture more details. The acquired images are stored in a lossless compression format, and each image is assigned a unique identification code, associated with the construction log for easy subsequent querying and analysis. The correction module 102 first applies a Gaussian filtering algorithm to the original image dataset for noise reduction, eliminating image noise caused by dust, vibration, and other factors in the construction environment. Then, it enhances image contrast using a histogram equalization method, making the borehole edges and platform structure clearer. Based on the positioning marker information, a perspective transformation matrix is ​​calculated to correct lens distortion. Finally, images acquired from different angles are spatially registered and mapped to the same coordinate system to generate standardized drilling platform images. The extraction module 103 applies the Canny edge detection algorithm to extract contours from standardized drilling platform images. This algorithm uses a dual-threshold method, with the lower threshold set to 40% of the higher threshold to ensure continuous and accurate detection. The extracted contours are then input into the Hough transform algorithm to identify straight lines and circular structural elements, locating the drilling platform boundaries and borehole positions. Based on matching feature points from images at different angles, point cloud data is constructed using structured light 3D reconstruction technology. Finally, through surface fitting and meshing, a 3D spatial model containing platform dimensions, borehole positions, and angles is established. The monitoring module 104 first sets the ideal state vector of the drilling trajectory based on the 3D spatial model of the drilling platform, including the borehole starting position, target depth, design angle, and direction. Real-time motion parameters of the drill rod are collected using drill rod position and depth sensors to construct observation vectors. An improved adaptive Kalman filter algorithm processes the observation vectors. This algorithm adds the function of real-time adjustment of the state noise covariance matrix based on geological parameters to the standard Kalman filter, adaptively correcting measurement errors caused by changes in geological conditions. Finally, the lateral and angular deviations between the filtered actual drilling trajectory and the ideal trajectory are calculated to generate drilling trajectory deviation data. The calculation module 105 applies the least squares method to fit the borehole centerline to the borehole trajectory deviation data, calculates the maximum distance from the actual borehole trajectory point to the fitted line, and determines the borehole straightness. The borehole depth is measured by analyzing the changes in drill rod feed length and drill bit position, and compared with the designed depth. The borehole diameter is measured at different depths to construct a borehole diameter variation curve. The gray-level co-occurrence matrix method is used to analyze the texture features of the borehole wall image, calculate texture parameters such as energy, entropy, and contrast, evaluate the borehole wall roughness, and form a borehole quality evaluation index.Analysis module 106 first monitors the anchor bolt installation process using a target tracking algorithm, recording changes in anchor bolt position, attitude, and grouting parameters to obtain monitoring data. The drilling quality evaluation indicators and the anchor bolt installation monitoring data are input into a hierarchical analysis model (AHP) to establish a quality evaluation system comprising four primary indicators: platform geometric parameters, drilling trajectory parameters, drilling quality parameters, and anchor bolt installation parameters. Entropy weighting is applied to assign weight coefficients to each indicator in the AHP model. This method automatically assigns weights based on the degree of indicator variation by calculating the information entropy of the indicators, avoiding the influence of subjective factors. Finally, based on the comprehensive quality score, four-level early warning thresholds are set, classifying the quality level into four grades: excellent, qualified, critical, and unqualified, generating construction quality levels and early warning information for the slope anchor bolt drilling platform.

[0024] In one specific embodiment, the acquisition module 101 is used for:

[0025] At least four positioning markers are set around the construction area of ​​the slope anchor drilling platform to establish a spatial coordinate system;

[0026] At least three 12-megapixel global shutter CCD sensor industrial cameras are deployed in the construction area of ​​the slope anchor drilling platform to form a multi-angle coverage network;

[0027] The high-resolution industrial camera continuously captures the construction process of the slope anchor drilling platform at a frequency of 5 frames per second, recording image timestamps and spatial location information.

[0028] The collected image data is losslessly compressed and stored, and a unique identification code is assigned to each image. An association index with the construction log is established to obtain the original image dataset with spatiotemporal tags.

[0029] Specifically, at least four positioning markers are set up around the construction area of ​​the slope anchor drilling platform to establish a spatial coordinate system for precise three-dimensional spatial positioning. These positioning markers are typically made of specially designed highly reflective materials, ensuring clear identification under different angles and lighting conditions. The four markers are arranged around the perimeter of the construction area, forming a three-dimensional reference system. Each marker has precise three-dimensional coordinate values, determined using a total station or RTK-GPS. This arrangement ensures that at least three markers appear simultaneously in the image from any angle, providing a reference for subsequent image geometric correction. At least three 12-megapixel global shutter CCD sensor industrial cameras are deployed around the construction area of ​​the slope anchor drilling platform to form a multi-angle coverage network for acquiring comprehensive construction information. The global shutter CCD sensor can simultaneously expose the entire sensor, avoiding distortion of moving objects in the image, making it particularly suitable for capturing high-speed moving parts during the drilling process. These cameras are typically arranged in a 120° angled triangular array to ensure monitoring of the drilling platform from different angles, eliminating occlusion problems caused by a single viewpoint. Each camera is equipped with a remote control unit to enable synchronized parameter adjustments.

[0030] Continuous acquisition of images at 5 frames per second using high-resolution industrial cameras during the construction of the slope anchor drilling platform is crucial for achieving precise spatiotemporal positioning. Timestamps are recorded with millisecond-level precision and synchronized with a global time server to ensure time consistency across multiple cameras. Spatial position information includes the camera's six-DOF attitude parameters (three position parameters and three angle parameters), acquired in real-time via an externally mounted inertial measurement unit. This high-frequency acquisition ensures the capture of instantaneous changes during drilling, particularly the critical moments of drill bit entry and exit from the rock. The acquired image data is losslessly compressed and stored, and each image is assigned a unique identification code. An index linking the image to the construction log is established, resulting in a spatiotemporally labeled raw image dataset, which forms the basis of data management. Lossless compression uses PNG format to ensure no loss of image quality. The unique identification code consists of the camera ID, acquisition timestamp, and serial number, in the form of "CAM01-20250519-153045-0001," ensuring that each image can be uniquely identified. The association with construction logs is achieved through time matching, linking images with construction parameters recorded at the same time, such as drilling depth, drilling pressure, and drilling speed, to form a complete spatiotemporal dataset, facilitating subsequent data mining and analysis. This data organization method solves the problems of data fragmentation and poor correlation in traditional construction monitoring.

[0031] In one specific embodiment, the correction module 102 is used for:

[0032] The original image dataset is subjected to Gaussian filtering algorithm for noise reduction to eliminate image noise caused by environmental factors;

[0033] The original image dataset after noise reduction is enhanced for contrast using histogram equalization to improve image detail.

[0034] Based on the positioning markers set around the construction area of ​​the slope anchor drilling platform, a perspective transformation matrix is ​​calculated to correct image distortion.

[0035] Spatial registration is performed on the corrected image to map images acquired from different angles to the same coordinate system, thereby generating the standardized drilling platform image.

[0036] Specifically, the correction module 102 applies a Gaussian filtering algorithm to the original image dataset for noise reduction, eliminating image noise caused by environmental factors. Gaussian filtering is a linear smoothing filtering algorithm that achieves image smoothing by convolving with a normal distribution (Gaussian distribution) function. In construction sites, factors such as dust, vibration, and changes in lighting can introduce noise into images, affecting subsequent image analysis. The Gaussian filtering algorithm uses a two-dimensional Gaussian function as the convolution kernel, performing a weighted average on each pixel and its neighborhood in the image. The weights are determined by the Gaussian function, with pixels closer to the center having larger weights. In practice, a 5×5 convolution kernel size and a standard deviation σ=1.5 parameter configuration are chosen. This configuration effectively removes salt-and-pepper noise and Gaussian noise commonly found in construction environments while preserving image edge details. The denoised original image dataset is then enhanced for contrast using histogram equalization to improve image detail. Histogram equalization is a non-linear image enhancement technique that redistributes image brightness values ​​to make the image histogram more uniformly distributed across the entire brightness range. In slope construction environments, lighting conditions are often unsatisfactory, leading to insufficient image contrast and difficulty in distinguishing borehole edges from platform structures. Histogram equalization calculates the cumulative distribution function of the image and then performs gray-level mapping, redistributing the gray values ​​of the original image to the range of 0-255. This brightens dark areas and appropriately compresses bright areas, thereby enhancing the local contrast of the image. Considering the characteristics of slope anchor drilling platform construction images, an adaptive histogram equalization method is adopted. The image is divided into multiple sub-blocks, each processed separately, and then the results are merged using bilinear interpolation. This effectively solves the problems of over-enhancement and noise amplification that may occur with global equalization.

[0037] Based on the positioning markers set around the construction area of ​​the slope anchor drilling platform, a perspective transformation matrix is ​​calculated to correct image distortion. Camera lens distortion can cause straight lines to appear as curves in the image, affecting the accuracy of geometric measurements. Using four pre-set positioning markers as references, a 3×3 perspective transformation matrix is ​​calculated by establishing the correspondence between the two-dimensional coordinates of the markers in the image and their actual three-dimensional spatial coordinates. This process first uses a feature matching algorithm to automatically detect the positions of the markers in the image, then solves for the parameters of the perspective transformation matrix using the least squares method, and finally applies this matrix to transform the entire image, correcting radial and tangential distortion. This step ensures that straight elements in the image (such as the edges of the drilling platform) retain their straightness after correction, laying the foundation for subsequent dimensional measurements.

[0038] Spatial registration is performed on the corrected images, mapping images acquired from different angles to the same coordinate system to generate a standardized drilling platform image. Spatial registration is a crucial step in solving multi-view image fusion, establishing the correspondence between images captured by different cameras. This process first extracts and matches feature points from images at different angles using feature point matching algorithms (such as SIFT or ORB), then uses the RANSAC algorithm to eliminate incorrect matches and calculates the relative poses between cameras. Based on this information, combined with the previously obtained perspective transformation matrix, all images are transformed to a unified "virtual camera" perspective, generating a complete, unobstructed standardized drilling platform image. This standardized image eliminates geometric distortion caused by perspective differences, ensuring measurement consistency in subsequent geometric feature extraction, and also providing accurate image correspondences for 3D reconstruction.

[0039] In one specific embodiment, the extraction module 103 is used for:

[0040] The Canny edge detection algorithm was applied to the standardized drilling platform image to extract the contour, and a dual threshold parameter was set, with the lower threshold being 40% of the higher threshold;

[0041] The extracted contours are input into the Hough transform algorithm to identify straight lines and circular structural elements, and to locate the boundary of the slope anchor drilling platform and the drilling position.

[0042] Based on the matching feature points of the standardized drilling platform images from different angles, point cloud data is constructed using structured light 3D reconstruction technology;

[0043] The point cloud data is subjected to surface fitting and meshing to establish a three-dimensional spatial model of the drilling platform, which includes the dimensions, drilling location, and drilling angle of the slope anchor drilling platform.

[0044] Specifically, the extraction module 103 first applies the Canny edge detection algorithm to the standardized drilling platform image to extract the contours, setting a dual threshold parameter, with the lower threshold being 40% of the higher threshold. Canny edge detection is a multi-stage edge detection algorithm, particularly suitable for extracting the contours of clearly structured engineering objects like drilling platforms. The algorithm first performs Gaussian filtering to smooth the image, then calculates the gradient magnitude and direction, followed by non-maximum suppression to obtain edges with a single pixel width, and finally connects the edges using a dual threshold method. In the application scenario of slope anchor drilling platforms, the higher threshold is typically set to the 70% quantile of the gradient magnitude, while the lower threshold is set to 40% of the higher threshold. This setting effectively suppresses noise interference while ensuring detection sensitivity. The algorithm outputs a binary edge image, clearly showing the platform contours and borehole boundaries, providing basic data for subsequent geometric feature recognition. The extracted contours are then input into the Hough transform algorithm to identify straight lines and circular structural elements, locating the slope anchor drilling platform boundaries and borehole positions. The Hough transform is a feature extraction technique capable of detecting parametric shapes in an image, such as straight lines and circles. For borehole platform boundary detection, the Hough linear transform is used to transform points in the binarized edge image to the Hough parameter space (ρ, θ), where ρ represents the perpendicular distance from the origin to the line, and θ represents the angle between the perpendicular line and the x-axis. By finding local maxima in the accumulator array in the parameter space, the main lines in the image are identified, which constitute the boundary of the borehole platform. For borehole location detection, the Hough circular transform is used to search in the three-dimensional parameter space (a, b, r), where (a, b) are the coordinates of the circle center, and r is the radius. The algorithm votes on the parameter space based on edge points, finds the parameters corresponding to local maxima in the accumulator array, and thus determines the location and diameter of the borehole. This method can effectively identify boreholes that are partially occluded or affected by lighting.

[0045] Based on matching feature points from standardized drilling platform images at different angles, point cloud data is constructed using structured light 3D reconstruction technology. Structured light 3D reconstruction is an active vision technology that recovers 3D information by projecting light of a known pattern onto a scene and then analyzing the deformation of the light pattern. In this module, using standardized images acquired from different angles, pixel-level correspondences are first established between image pairs using feature matching algorithms (such as SIFT), identifying the projection position of the same physical point in different images. Then, based on the camera's intrinsic and extrinsic parameters and the established correspondences, the coordinates of the point in 3D space are calculated using the principle of triangulation. For complex objects such as slope anchor drilling platforms, a multi-view fusion strategy is adopted. Point cloud data from different perspectives are registered and fused using the Iterative Closest Point (ICP) algorithm to eliminate blind spots and obtain dense point cloud data covering the entire platform. These point clouds accurately describe the platform's geometry and surface features.

[0046] Surface fitting and meshing were performed on point cloud data to establish a 3D spatial model of the drilling platform, including its dimensions, drilling locations, and drilling angles. Point cloud data inherently contains noise and outliers, requiring processing to construct a high-quality 3D model. First, the point cloud was denoised and downsampled. A statistical outlier filtering algorithm was used to remove noise points, followed by voxel mesh filtering for uniform downsampling, reducing data volume while preserving geometric details. Next, surface reconstruction was performed. The RANSAC plane fitting algorithm was applied to extract the main planes of the platform, and a cylinder fitting algorithm was applied to identify the drilling axis and diameter of the boreholes. Finally, a complete triangular mesh model was generated using the Poisson surface reconstruction algorithm, and mesh optimization processes, such as mesh simplification and smoothing, were performed to form an accurate and efficient 3D spatial model. This model accurately records the platform's length, width, height, and other dimensional parameters, the spatial coordinates of the boreholes, and key geometric information such as the angle between the borehole axis and the horizontal plane, laying the foundation for subsequent borehole trajectory monitoring.

[0047] In one specific embodiment, the monitoring module 104 is used for:

[0048] The ideal state vector of the drilling trajectory is set according to the three-dimensional spatial model of the drilling platform, including the starting position of the drilling, the target depth, the design angle and the direction;

[0049] Real-time motion parameters of the drill pipe are collected by drill pipe position and depth sensors to construct an observation vector.

[0050] The observation vector is processed by the improved adaptive Kalman filter algorithm, wherein the state noise covariance matrix is ​​adjusted in real time according to geological parameters;

[0051] The lateral and angular deviations between the filtered actual drilling trajectory and the ideal trajectory are calculated to generate the borehole trajectory deviation data.

[0052] Specifically, the monitoring module 104 first sets the ideal state vector of the drilling trajectory based on the three-dimensional spatial model of the drilling platform, including the starting position of the drilling, the target depth, the design angle, and the direction. The ideal state vector is a mathematical expression of the drilling design parameters and is used as a benchmark for comparison with the actual trajectory. In slope anchor drilling projects, the ideal trajectory of each borehole is a straight line from the starting point to the target point. The starting position is determined by the surface coordinates of the drilling platform extracted from the three-dimensional spatial model, usually represented in a global coordinate system. The design angle includes the angle between the borehole and the horizontal plane and the angle between the borehole's projection on the horizontal plane and the north direction; these two angles together determine the spatial direction of the borehole. The target depth is the distance from the starting point along the borehole axis. Based on these parameters, the ideal state vector is constructed as a six-dimensional vector, describing the ideal spatial trajectory that the borehole should follow under the design requirements. Real-time motion parameters of the drill rod are collected by drill rod position sensors and depth sensors to construct the observation vector. Drill rod position sensors typically include tilt sensors and azimuth sensors installed on the drilling rig, which measure the angle between the drill rod and the vertical direction and the direction of the drill rod's projection on the horizontal plane, respectively. Depth sensors monitor the feed length of the drill pipe, typically through encoders or displacement sensors. These sensors acquire data at a high frequency to ensure the capture of minute changes during drilling. The observation vector consists of these real-time measurements, representing the current drill bit position coordinates, the current drill pipe inclination and azimuth angles, and the current drilling depth. Each parameter in the observation vector contains measurement noise, which needs to be processed using filtering algorithms to obtain accurate trajectory information.

[0053] The observation vectors are processed using an improved adaptive Kalman filter algorithm, where the state noise covariance matrix is ​​adjusted in real time based on geological parameters. The standard Kalman filter algorithm assumes that the system and measurement noise are constant, but in actual drilling, noise characteristics change due to variations in geological conditions. The improved adaptive Kalman filter algorithm introduces a dynamic adjustment mechanism for the covariance matrix, adaptively updating it based on geological parameters and real-time feedback. The algorithm first establishes state transition equations and observation equations; the former describes the dynamic changes in drill bit position and attitude, while the latter describes the relationship between sensor measurements and the actual state. Then, it performs two main steps: prediction and updating. The prediction step predicts the current state based on the previous state and the state transition equation; the updating step corrects the prediction results based on observation data. The improvement lies in the introduction of geological parameter influence factors. When the drill bit encounters hard rock formations, the state noise covariance is increased to accommodate possible trajectory deviations; when drilling into soft rock formations, the covariance is decreased to improve filtering accuracy. This adaptive mechanism significantly improves the accuracy of borehole trajectory monitoring under complex geological conditions.

[0054] The lateral and angular deviations between the filtered actual drilling trajectory and the ideal trajectory are calculated to generate borehole trajectory deviation data. Lateral deviation is the shortest distance from the actual borehole trajectory point to the ideal trajectory line, calculated using the point-to-line distance formula. During drilling, the lateral deviation value is recorded every 10 cm of drilling depth to construct a complete deviation curve. Angular deviation is the angle between the actual drilling direction and the design direction. It is obtained by fitting the current drilling direction vector through three consecutive drilling points, and then calculating the angle with the design direction vector. The borehole trajectory deviation data contains three key sets of information: depth, lateral deviation, and angular deviation, forming a complete borehole quality monitoring dataset. This dataset not only reflects the current quality status of the borehole but also predicts the final borehole quality through trend analysis, providing a basis for timely adjustment of drilling parameters. This real-time monitoring and analysis effectively solves the problem of delayed detection of quality problems in traditional borehole construction, significantly improving the construction quality and efficiency of slope anchor drilling.

[0055] In one specific embodiment, the calculation module 105 is used for:

[0056] The least squares method is applied to fit the borehole centerline to the borehole trajectory deviation data, the maximum distance from the actual borehole trajectory point to the fitted line is calculated, and the borehole straightness is determined.

[0057] The drilling depth is measured by analyzing the changes in drill pipe feed length and drill bit position, and then compared with the designed depth.

[0058] Based on the edge detection results, the borehole diameter is measured at different depth positions to construct a borehole diameter variation curve;

[0059] The texture features of the borehole wall image are analyzed using the gray-level co-occurrence matrix method. Texture parameters, including energy, entropy, and contrast, are calculated to evaluate the roughness of the borehole wall and form the borehole quality evaluation index.

[0060] Specifically, the calculation module 105 first applies the least squares method to fit the borehole centerline to the borehole trajectory deviation data, calculates the maximum distance from the actual borehole trajectory points to the fitted line, and determines the borehole straightness. The least squares method is a mathematical optimization technique that finds the best function match for the data by minimizing the sum of squared errors. During the borehole centerline fitting process, the spatial point set in the borehole trajectory deviation data is used as input, and the objective function is constructed as the sum of squared distances from all points to the line to be fitted. By solving for the straight line parameters that minimize this objective function, the optimal borehole centerline is obtained. After fitting, the distance from each point on the actual borehole trajectory to the fitted centerline is calculated, and the maximum value is taken as the borehole straightness index. Borehole straightness directly reflects the curvature of the borehole trajectory and is an important parameter for evaluating borehole quality. The borehole depth is measured by analyzing the drill rod feed length and drill bit position changes, and compared with the designed depth. The drill rod feed length is directly measured by a displacement sensor installed on the drilling rig, recording the distance the drill rod moves from its initial position. The drill bit position change is acquired by a position sensor in the aforementioned monitoring module, recording the real-time coordinates of the drill bit in three-dimensional space. The drilling depth calculation comprehensively considers both data sources. By comparing the projected distance of the drill pipe feed length and the three-dimensional position change of the drill bit, errors that may be caused by drill pipe bending or extension are eliminated. The calculated actual drilling depth is compared with the target depth specified in the design documents to evaluate the depth error, which serves as the basis for evaluating drilling quality.

[0061] Based on edge detection results, the borehole diameter is measured at different depths to construct a borehole diameter variation curve. This step utilizes the edge detection images obtained from the extraction module and processes them for different depth sections. For each depth section image, the Hough circle transform algorithm is applied to identify the borehole boundary, calculate the center position and radius, and thus obtain the borehole diameter at that depth. To ensure measurement accuracy, multi-angle images are acquired at each depth location, and the results of multi-view measurements are combined to eliminate occlusion and errors that may be caused by a single viewpoint. By measuring the diameter value at fixed intervals (usually 20 cm) along the entire length of the borehole, a curve reflecting the change of borehole diameter with depth is constructed. This curve can clearly show whether there are quality problems such as taper, necking, or enlargement in the borehole.

[0062] The gray-level co-occurrence matrix (GLCM) method was used to analyze the texture features of borehole wall images, calculating texture parameters including energy, entropy, and contrast to evaluate borehole wall roughness and form a borehole quality evaluation index. The GLCM is a statistical method for describing image texture features. It constructs a matrix reflecting texture information by calculating the co-occurrence frequency of gray-level values ​​of pixel pairs with specific positional relationships in the image. For borehole wall images, the distance and directional relationships between pixel pairs are first defined, typically selecting four directions (vertical, horizontal, and diagonal) and a distance range of 1-3 pixels, and the corresponding GLCM is calculated. Based on the GLCM, second-order statistical features such as energy, entropy, and contrast are calculated. Energy reflects the uniformity of the texture; a higher value indicates a more uniform texture. Entropy reflects the complexity of the texture; a higher value indicates a more complex texture. Contrast reflects the clarity of the texture; a higher value indicates a more distinct texture boundary. These parameters comprehensively evaluate the roughness of the borehole wall. Excessive roughness may affect the bonding effect between the anchor bolt and the borehole wall, while excessively low roughness may lead to poor grouting results.

[0063] In one specific embodiment, the analysis module 106 is used for:

[0064] The anchor bolt installation process is monitored using a target tracking algorithm to record changes in anchor bolt position, attitude, and grouting parameters, thereby obtaining monitoring data for the anchor bolt installation process.

[0065] The drilling quality evaluation indicators and the monitoring data of the anchor bolt installation process are input into the hierarchical analysis model to establish a quality evaluation system.

[0066] The entropy weight method is applied to assign weight coefficients to each indicator in the hierarchical analysis model to calculate the comprehensive quality score of each anchor borehole.

[0067] Based on the comprehensive quality score, four warning thresholds are set to divide the quality level into four levels: excellent, qualified, critical, and unqualified, thereby generating the construction quality level and warning information of the slope anchor drilling platform.

[0068] Specifically, the analysis module 106 first monitors the anchor bolt installation process using a target tracking algorithm, recording changes in anchor bolt position, attitude, and grouting parameters to obtain monitoring data. The target tracking algorithm is a technique in computer vision used to track moving targets; in anchor bolt installation monitoring, a feature-point-based tracking method is employed. This method first identifies distinctive markers on the anchor bolt, such as the connection between the head and tail of the anchor bolt, and the boundary of the anti-corrosion layer, and then tracks the movement trajectory of these feature points in a continuous image sequence. By analyzing the spatial changes of these feature points, the moving speed, insertion depth, and attitude angle of the anchor bolt are calculated. Simultaneously, image recognition technology is used to read the pressure gauge and flow rate count values ​​on the grouting equipment to obtain parameters such as grouting pressure, grouting volume, and grouting speed. This data is recorded in time-series format, forming a complete anchor bolt installation process monitoring dataset that comprehensively reflects the quality status of the anchor bolt installation. The drilling quality evaluation indicators and the anchor bolt installation process monitoring data are input into a hierarchical analysis model (AHP) to establish a quality evaluation system. The AHP is a multi-criteria decision-making method that simplifies the decision-making process by decomposing complex problems into a hierarchical structure. In the quality evaluation of slope anchor drilling platforms, this model constructs a three-layer structure: the top layer represents the overall objective, namely, a comprehensive evaluation of construction quality; the middle layer comprises primary indicators, including four main categories: platform geometric parameters, drilling trajectory parameters, drilling quality parameters, and anchor installation parameters; the bottom layer consists of secondary indicators, detailing the specific evaluation items for each type of parameter. Platform geometric parameters include platform levelness and stability; drilling trajectory parameters include straightness and angular deviation; drilling quality parameters include depth, diameter, and surface roughness; and anchor installation parameters include anchor position accuracy, grouting fullness, and anchoring force. Each indicator has clear quantitative standards and judgment rules, forming a comprehensive quality evaluation system.

[0069] Entropy weighting is applied to assign weight coefficients to each indicator in the analytic hierarchy process (AHP) model, calculating the comprehensive quality score for each anchor borehole. Entropy weighting is an objective weighting method based on information entropy theory, determining weights by calculating the information entropy of the indicators. Higher information entropy indicates greater uncertainty in the indicator, less impact on decision-making, and consequently, a smaller weight. The specific process includes: first, standardizing the values ​​of each indicator to eliminate the influence of dimensions; then calculating the information entropy of each indicator, taking into account its distribution across all evaluation objects; next, calculating the entropy weight based on the information entropy, which is inversely proportional to the information entropy; finally, combining the AHP structure, the entropy weights of the secondary indicators are aggregated in a bottom-up order to form the weights of the primary indicators, and the entropy weights of the primary indicators are then aggregated to form the overall evaluation weight allocation. Based on the determined weight coefficients and standardized indicator values, a weighted summation is used to calculate the comprehensive quality score for each anchor borehole, which comprehensively reflects the overall quality level of the borehole and anchor installation.

[0070] Based on a comprehensive quality score, a four-level early warning threshold is set, classifying quality levels into four grades: excellent, qualified, critical, and unqualified. This generates construction quality levels and early warning information for slope anchor drilling platforms. The early warning thresholds are set based on extensive engineering practice data and quality standard requirements. Generally, scores above 90 are defined as excellent, 75-89 as qualified, 60-74 as critical, and below 60 as unqualified. For different grades, the system generates corresponding early warning information: excellent requires no special handling; qualified provides routine acceptance suggestions; critical issues a yellow warning, indicating the need for close attention and necessary reinforcement measures; unqualified issues a red warning, requiring the cessation of subsequent procedures and rework or reinforcement. The early warning information also includes specific problem descriptions and handling suggestions, such as suggesting back-drilling for insufficient straightness or supplementing grouting for incomplete grouting. The system also generates a quality distribution map, visually displaying the quality status of each borehole in the project, helping construction personnel quickly identify problem areas and conduct targeted quality control.

[0071] The above embodiments are only used to illustrate the technical solutions of the present invention, and are not intended to limit it. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A machine vision-based intelligent inspection system for the construction quality of slope anchor drilling platforms, characterized in that the system... include: The acquisition module is used to acquire multi-angle images of the construction area of ​​the slope anchor drilling platform using a high-resolution industrial camera, and obtain a raw image dataset with spatiotemporal tags. The correction module is used to perform image preprocessing and geometric correction on the original image dataset to obtain a standardized drilling platform image; The extraction module is used to extract the geometric feature parameters of the drilling platform based on the standardized drilling platform image. Specifically, it is used to: apply the Canny edge detection algorithm to the standardized drilling platform image to extract the contour, and set dual threshold parameters, with the lower threshold being 40% of the higher threshold. The extracted contours are input into the Hough transform algorithm to identify straight lines and circular structural elements, and to locate the boundary and drilling position of the slope anchor drilling platform. Based on the matching feature points of the standardized drilling platform images at different angles, point cloud data is constructed using structured light 3D reconstruction technology. The point cloud data is then subjected to surface fitting and meshing to establish a 3D spatial model of the drilling platform that includes the dimensions, drilling position, and drilling angle of the slope anchor drilling platform. The monitoring module is used to apply an improved adaptive Kalman filter algorithm to the three-dimensional spatial model of the drilling platform for real-time monitoring of the drilling trajectory, and to obtain drilling trajectory deviation data. Specifically, it is used for: setting an ideal state vector of the drilling trajectory based on the three-dimensional spatial model of the drilling platform, including the starting position of the drilling, the target depth, the design angle, and the direction; collecting real-time motion parameters of the drill rod through the drill rod position sensor and the depth sensor to construct an observation vector; processing the observation vector through the improved adaptive Kalman filter algorithm, wherein the state noise covariance matrix is ​​adjusted in real time according to geological parameters, and the improved adaptive Kalman filter algorithm introduces a dynamic adjustment mechanism for the covariance matrix; calculating the lateral deviation and angular deviation between the filtered actual drilling trajectory and the ideal trajectory to generate the drilling trajectory deviation data. The calculation module is used to calculate key parameters of borehole quality based on the borehole trajectory deviation data, and obtain borehole quality evaluation indicators. The analysis module is used to comprehensively analyze the drilling quality evaluation indicators and the monitoring data of the anchor bolt installation process to obtain the construction quality level and early warning information of the slope anchor bolt drilling platform.

2. The intelligent inspection system for construction quality of slope anchor drilling platform based on machine vision as described in claim 1, characterized in that, The acquisition module is used for: At least four positioning markers are set around the construction area of ​​the slope anchor drilling platform to establish a spatial coordinate system; At least three 12-megapixel global shutter CCD sensor industrial cameras are deployed in the construction area of ​​the slope anchor drilling platform to form a multi-angle coverage network; The high-resolution industrial camera continuously captures the construction process of the slope anchor drilling platform at a frequency of 5 frames per second, recording image timestamps and spatial location information. The collected image data is losslessly compressed and stored, and a unique identification code is assigned to each image. An association index with the construction log is established to obtain the original image dataset with spatiotemporal tags.

3. The intelligent inspection system for construction quality of slope anchor drilling platform based on machine vision as described in claim 1, characterized in that, The correction module is used for: The original image dataset is subjected to Gaussian filtering algorithm for noise reduction to eliminate image noise caused by environmental factors; The original image dataset after noise reduction is enhanced for contrast using histogram equalization to improve image detail. Based on the positioning markers set around the construction area of ​​the slope anchor drilling platform, a perspective transformation matrix is ​​calculated to correct image distortion. Spatial registration is performed on the corrected images to map images acquired from different angles to the same coordinate system, thereby generating the standardized drilling platform image.

4. The intelligent inspection system for construction quality of slope anchor drilling platform based on machine vision as described in claim 1, characterized in that, The computing module is used for: The least squares method is applied to fit the borehole centerline to the borehole trajectory deviation data, the maximum distance from the actual borehole trajectory point to the fitted line is calculated, and the borehole straightness is determined. The drilling depth is measured by analyzing the changes in drill pipe feed length and drill bit position, and then compared with the designed depth. Based on the edge detection results, the borehole diameter is measured at different depth positions to construct a borehole diameter variation curve; The texture features of the borehole wall image are analyzed using the gray-level co-occurrence matrix method. Texture parameters, including energy, entropy, and contrast, are calculated to evaluate the roughness of the borehole wall and form the borehole quality evaluation index.

5. The intelligent inspection system for construction quality of slope anchor drilling platform based on machine vision as described in claim 1, characterized in that, The analysis module is used for: The anchor bolt installation process is monitored using a target tracking algorithm to record changes in anchor bolt position, attitude, and grouting parameters, thereby obtaining monitoring data for the anchor bolt installation process. The drilling quality evaluation indicators and the monitoring data of the anchor bolt installation process are input into the hierarchical analysis model to establish a quality evaluation system. The entropy weight method is applied to assign weight coefficients to each indicator in the hierarchical analysis model to calculate the comprehensive quality score of each anchor borehole. Based on the comprehensive quality score, four warning thresholds are set to divide the quality level into four levels: excellent, qualified, critical, and unqualified, thereby generating the construction quality level and warning information of the slope anchor drilling platform.

Images