Highway construction inspection optimization method based on artificial intelligence

Abstract

The invention relates to the technical field of construction inspection, and discloses a highway construction inspection optimization method based on artificial intelligence, and the method comprises the steps: recognizing a foundation pit boundary pixel in a construction ground image through a boundary recognition model, mapping the foundation pit boundary pixel to a highway construction site, and carrying out the construction of the highway. Taking an area surrounded by the pixel mapping position of the foundation pit boundary as a foundation pit; parallax information of the foundation pit is obtained through calculation based on the left view and the right view of the foundation pit, depth and width estimation is conducted on the foundation pit through a stereo matching algorithm, and the depth and the width of the foundation pit are obtained; and generating a safety risk value of the discrete grid based on the foundation pit information of the foundation pit in the discrete grid, and performing secondary inspection on the discrete grid with the safety risk value higher than an allowable risk. According to the invention, the depth, width and area of the foundation pit are obtained through the stereo matching algorithm, and the discrete grid is combined to generate the safety risk value, so that precise secondary inspection of the high-risk discrete grid and refined safety management of highway construction are realized.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of construction inspection, in particular to an image recognition technology, and more particularly to a highway construction inspection optimization method based on artificial intelligence. BACKGROUND

[0002] Highway construction and maintenance are important basic links to ensure long-term safe operation of roads and improve traffic efficiency. With the continuous growth of traffic flow and the continuous improvement of road use frequency, construction operations have gradually shifted from traditional manual inspection to intelligent and information-based direction. However, in the process of construction and maintenance, foundation pits, as a common construction structure, have non-negligible safety risks. On the one hand, foundation pits are usually formed by excavation operations, and their boundaries and forms are often complex and variable. If they are not monitored and accurately labeled in real time, it is easy to cause construction machinery to misenter or workers to fall, causing serious personal injury and equipment damage. On the other hand, if the foundation pit is not identified in time or the parameter evaluation is not accurate, vehicles driving near the construction area may be affected by problems such as foundation pit collapse and insufficient road support, thus causing traffic accidents and posing a hidden danger to road safety and efficiency.

[0003] Foundation pits not only differ in depth and width, but also have dynamic adjustments in area and spatial distribution position with the change of construction progress. Therefore, the traditional method relying on manual inspection and experience judgment cannot guarantee comprehensive understanding and accurate updating of foundation pit information, and has problems of monitoring delay and insufficient recognition accuracy. In actual construction scenarios, night work, complex terrain, and rainwater accumulation and other environmental factors also make the boundary recognition and risk determination of foundation pits more difficult. Therefore, how to realize rapid identification and high-precision measurement of foundation pit areas has become a key technical bottleneck restricting construction safety management.

[0004] Existing researches are mostly focused on highway inspection methods based on artificial intelligence and big data. The patent CN117726324B, a highway traffic construction inspection method and system based on data recognition, proposes an inspection method combining neural networks and prediction models. This method obtains the maintenance value of the future time period, predicts vehicle transportation data, temperature data, and precipitation data, and imports these data into the constructed neural network model to obtain the trend of future maintenance value. Further, by estimating the time when the maintenance value reaches the maintenance threshold, the road maintenance time is predicted and an alarm is realized, thereby reducing the time difference between road damage and maintenance and improving the foresight and timeliness of road maintenance. Although the above patent has made progress in road damage prediction and maintenance time warning, it still has the following technical difficulties: the patent focuses on time dimension prediction and cannot quantitatively measure the spatial geometric features such as the boundary, depth, and width of the foundation pit; the alarm is given by predicting the maintenance time, and the actual topographic information and structural risks of the construction site are not combined for targeted intelligent analysis of the most dangerous foundation pit in construction.

[0005] To solve the problem, the present application provides a highway construction inspection optimization method based on artificial intelligence, which automatically calculates and measures the foundation pit at the construction site by introducing artificial intelligence technology, realizes intelligent inspection, improves the efficiency and accuracy of inspection, reduces the cost of manual inspection, and improves the overall management level of the construction process. SUMMARY

[0006] The present application provides a highway construction inspection optimization method based on artificial intelligence. Traditional construction inspection relies on manual visual inspection or single camera equipment, which is easily disturbed by changes in light, dust obstruction, and weak texture, resulting in inaccurate detection of foundation pit boundaries. The S1 step of the present application uses a fixed position depth camera combined with a boundary recognition model to automatically extract and map the foundation pit boundary pixels, ensuring accurate identification of the foundation pit area in complex environments and reducing false positives and false negatives. In the S3 step, the stereo matching algorithm is constructed by calculating the disparity of the binocular camera and introducing the three-dimensional point cloud of the mobile inspection vehicle as a geometric constraint, greatly improving the accuracy and robustness of the foundation pit depth and width estimation. Traditional construction safety assessment is mostly based on single-point measurement or experience-based judgment, lacking detailed spatial distribution analysis and easily ignoring local high-risk areas. In the S4 step, the construction site is divided into discrete grid areas, and the safety risk value of each discrete grid is quantitatively calculated based on comprehensive information such as foundation pit area, depth, and width, realizing the spatialization and visualization of risk and effectively identifying high-risk areas.

[0007] To achieve the above purpose, the present application provides a highway construction inspection optimization method based on artificial intelligence, comprising the following steps: S1: Collect the construction ground image of the highway construction site using a fixed-position depth camera, identify the foundation pit boundary pixels in the construction ground image using a boundary recognition model, map the foundation pit boundary pixels to the highway construction site, and take the area surrounded by the mapping position of the foundation pit boundary pixels as the foundation pit; S2: Control the laser radar scanner deployed on the mobile inspection vehicle to collect auxiliary point cloud data of the foundation pit, and control the binocular camera deployed on the mobile inspection vehicle to collect the left view and the right view of the foundation pit; S3: Calculate the parallax information of the foundation pit based on the left view and the right view, and combine the auxiliary point cloud data to estimate the depth and width of the foundation pit using a stereo matching algorithm, to obtain the depth and width of the foundation pit; S4: Divide the highway construction site into discrete grid areas, generate safety risk values for the discrete grid areas based on the foundation pit information of the foundation pit in the discrete grid, and perform secondary inspection on the discrete grid areas with safety risk values higher than the allowable risk, wherein the foundation pit information includes the area, depth and width of the foundation pit.

[0008] As a further improved method of the present application: Further, the boundary recognition model is used to identify the foundation pit boundary pixels in the construction ground image, comprising: A plurality of fixed-position depth cameras are deployed on the highway construction site, and the construction ground image of the highway construction site is collected using the deployed cameras, and the collected construction ground image is sent to the boundary recognition model; The boundary recognition model includes an input layer, a contrast enhancement layer and a boundary recognition layer, the input layer is used to receive the construction ground image and perform grayscale processing to obtain a construction ground grayscale image, the contrast enhancement layer is used to calculate the contrast of the construction ground grayscale image, the histogram equalization method is used for contrast enhancement processing on the construction ground grayscale image with low contrast, and gamma correction processing is performed on the overexposed or underexposed construction ground grayscale image to obtain a construction ground contrast enhancement image, the boundary recognition layer is used to identify the foundation pit boundary pixels in the construction ground contrast enhancement image, and the pixel coordinates of the foundation pit boundary pixels are extracted; The boundary recognition layer uses an improved YOLOv8 model structure, and the improvement method is: introducing a multi-scale feature fusion structure in the backbone network; introducing a spatial attention mechanism in the detection head part; The boundary recognition model is used to identify the boundary of the construction ground image, and the pixel coordinates of the foundation pit boundary pixels are obtained, and the pixel coordinates of the foundation pit boundary pixels are used to map the foundation pit boundary pixels to the highway construction site.

[0009] Furthermore, the step of mapping the foundation pit boundary pixels to the highway construction site using the pixel coordinates of the foundation pit boundary pixels, and defining the area enclosed by the mapped positions of the foundation pit boundary pixels as the foundation pit, includes: Extract the depth value at the pixel coordinates of the pixels at the edge of the pit, and combine the depth value to map the pixel coordinates to the camera coordinate system based on the intrinsic parameter matrix of the depth camera, so as to obtain the three-dimensional camera coordinates of the pixel coordinates of the pixels at the edge of the pit in the camera coordinate system. The 3D camera coordinates are transformed to a unified highway coordinate system using the extrinsic rotation matrix and extrinsic translation vector of the depth camera, thus obtaining the 3D highway coordinates of the 3D camera coordinates in the highway coordinate system. Connect the three-dimensional highway coordinates in the highway coordinate system to form a closed polygon. The closed polygon is the area enclosed by the pixel mapping position of the pit boundary.

[0010] Furthermore, the system controls the lidar scanner deployed on the mobile inspection vehicle to collect auxiliary point cloud data of the foundation pit, and controls the binocular camera deployed on the mobile inspection vehicle to collect left and right views of the foundation pit, including: At highway construction sites, mobile inspection vehicles are deployed in appropriate locations around the foundation pits to ensure that the mobile inspection vehicles can cover each foundation pit along the inspection route. The mobile inspection vehicle is equipped with a lidar scanner and a binocular camera. The binocular camera includes a left-view camera and a right-view camera, which are used to acquire the left view and the right view, respectively. The mobile inspection vehicle is controlled to move along the perimeter of the foundation pit, so that the laser radar scanner can perform a 360° rotation scan of the foundation pit at a preset scanning frequency, and collect three-dimensional point clouds of the foundation pit and its perimeter, forming a three-dimensional point cloud set as auxiliary point cloud data of the foundation pit. The three-dimensional point cloud is in the form of three-dimensional coordinates. The mobile inspection vehicle is controlled to simultaneously activate the left and right view cameras to capture images, obtaining a panoramic view of the foundation pit from both the left and right sides.

[0011] Set the parallax search range to ; Calculate the disparity values ​​of any pixel coordinates in the grayscale left view under different disparities, and use them as the disparity information of the foundation pit: ; in, This represents the disparity value of pixel coordinate u in the grayscale left view under disparity q. This represents the set of pixel coordinates of the grayscale left view. This represents the left view in grayscale centered at pixel coordinate u. Pixel region, v represents pixel region Any pixel coordinate in the array, represents a gray value of the pixel coordinate v in the grayed left view, represents a pixel coordinate after the pixel coordinate v in the horizontal direction is shifted by q pixels, represents a gray value of the pixel coordinate v in the grayed right view; represents a gray value of the pixel coordinate v in the grayed left view, obtain auxiliary point cloud data of the foundation pit.

[0012] Further, in combination with the auxiliary point cloud data, a stereo matching algorithm is used to estimate the depth and width of the foundation pit, to obtain the depth and width of the foundation pit, including: calculate the point cloud consistent matching cost between any pixel coordinate in the grayed left view and the auxiliary point cloud data, wherein the point cloud consistent matching cost is the minimum coordinate difference between the pixel coordinate and the three-dimensional point cloud projection result of all three-dimensional point clouds in the auxiliary point cloud data, and the three-dimensional point cloud projection result is the projection of the three-dimensional point cloud to the pixel coordinate system in which the pixel coordinate is located; add the point cloud consistent matching cost between the pixel coordinate in the grayed left view and the auxiliary point cloud data and the disparity value of the pixel coordinate at different disparities, as the initial matching cost of the pixel coordinate in the grayed left view at different disparities; accumulate the initial matching cost of the pixel coordinate in the grayed left view at different disparities in different directions, and sum the cost accumulation results in multiple directions as the path accumulation cost of the pixel coordinate at different disparities; select the disparity that minimizes the path accumulation cost of the pixel coordinate in the grayed left view, generate the disparity depth of the pixel coordinate in the grayed left view based on the selected disparity, the camera baseline and the focal length, take the disparity depth as the depth value of the pixel coordinate in the grayed left view, map the pixel coordinate in the grayed left view to the highway coordinate system in the manner described in step S1, obtain the mapped three-dimensional highway coordinate, and take the coordinate value of the mapped three-dimensional highway coordinate in the Z axis as the depth of the pixel coordinate in the grayed left view at the foundation pit position; select the mapped three-dimensional highway coordinates of all pixel coordinates in the grayed left view, project the mapped three-dimensional highway coordinates into a plane coordinate system in the highway coordinate system, obtain the projection coordinates of each mapped three-dimensional working coordinate, calculate the Euclidean distance between any two projection coordinates, and select the maximum Euclidean distance as the width of the foundation pit; select the maximum value of the depth of all pixel coordinates in the grayed left view at the foundation pit position as the depth of the foundation pit.

[0013] Further, step S4 includes: project the highway construction site to a plane coordinate system in the highway coordinate system; The projected highway construction site is divided into a plurality of square discrete grids; The safety risk value of the discrete grid is generated based on the pit information of the pits in the discrete grid.

[0014] Further, the safety risk value calculation formula of the discrete grid is: ; Wherein, R represents the safety risk value of the discrete grid, Len represents the side length of the discrete grid, H represents the number of pits in the discrete grid, represents the area of the hth pit in the discrete grid, represents the depth of the hth pit in the discrete grid, represents the width of the hth pit in the discrete grid, represents the maximum depth of all pits in the highway construction site, The normalized values of the area , the depth and the width are sequentially respectively, represents the volume proportion factor of the hth pit; Both represent the effect coefficient, represents the nonlinear amplification coefficient.

[0015] Compared with the prior art, the present application proposes a highway construction inspection optimization method based on artificial intelligence, which has the following beneficial effects: Firstly, the present application improves the YOLOv8 model to realize the accurate identification of the pit boundary pixels in the construction ground contrast enhancement image, significantly improves the automation and precision of the construction site monitoring; Specifically, by introducing a multi-scale feature fusion structure into the backbone network, the model can capture the overall pit contour at low resolution and the long boundary line and auxiliary marker information at medium and high resolution at the same time, thereby effectively solving the problem that the traditional single-scale feature extraction cannot accurately identify complex edges; The introduction of spatial attention mechanism makes the model produce higher response to key boundaries and weak texture areas in the fused feature map, and improves the robustness under the conditions of light change, shadow shielding and noise interference.

[0016] Meanwhile, this invention tightly couples the discrete grid of the highway construction site with the information of the foundation pit, significantly improving the precision of construction risk assessment. Specifically, the formula introduces normalization and nonlinear penalty terms on the basis of traditional area, depth, and width indicators. This makes the safety risk value amplified for foundation pits with excessive depth or width exceeding the limit, which is more in line with the actual propagation law of construction safety hazards. By coupling the discrete grid scale with area proportion factors and volume proportion factors, it can reflect the degree of damage of the foundation pit to the overall stability of the local foundation, avoiding the shortcomings of simple area or depth evaluation methods in missing large-scale hazards. In this way, not only can the comparable calculation of foundation pits of different sizes be achieved, but also the discrete grids that exceed the allowable risk can be highlighted, thereby providing a scientific basis and accurate positioning for subsequent secondary inspections, effectively improving the intelligence and safety assurance level of highway construction site inspections. Attached Figure Description

[0017] Figure 1 This is a flowchart illustrating an artificial intelligence-based optimization method for highway construction inspection, as provided in an embodiment of the present invention.

[0018] Figure 2 This is a flowchart of the foundation pit boundary pixel recognition process provided in an embodiment of the present invention. Detailed Implementation

[0019] The realization of the objectives, functional characteristics, and advantages of this invention will be further explained in conjunction with the embodiments and with reference to the accompanying drawings. It should be understood that the specific embodiments described herein are merely illustrative of the invention and are not intended to limit the invention.

[0020] This invention provides an AI-based optimization method for highway construction inspection. The executing entity of this AI-based highway construction inspection optimization method includes, but is not limited to, at least one of the following electronic devices that can be configured to execute the method provided in this invention: a server, a terminal, etc. In other words, the AI-based highway construction inspection optimization method can be executed by software or hardware installed on a terminal device or a server device, and the software can be a blockchain platform. The server includes, but is not limited to, a single server, a server cluster, a cloud server, or a cloud server cluster.

[0021] Reference Figure 1 as well as Figure 2 Embodiment 1 of the present invention is as follows: An AI-based optimization method for highway construction inspection includes the following steps: S1: collecting a construction ground image of a highway construction site by using a fixed-position depth camera, identifying a foundation pit boundary pixel in the construction ground image by using a boundary identification model, mapping the foundation pit boundary pixel to the highway construction site, and taking a region surrounded by a mapping position of the foundation pit boundary pixel as a foundation pit.

[0022] identifying a foundation pit boundary pixel in the construction ground image by using the boundary identification model, comprising: deploying a plurality of fixed-position depth cameras in the highway construction site, collecting a construction ground image of the highway construction site by using the deployed cameras, and sending the collected construction ground image to the boundary identification model; The boundary identification model comprises an input layer, a contrast enhancement layer, and a boundary identification layer. The input layer is used to receive the construction ground image and perform grayscale processing to obtain a construction ground grayscale image. The contrast enhancement layer is used to calculate the contrast of the construction ground grayscale image. The histogram equalization method is used to enhance the contrast of the construction ground grayscale image with low contrast. Gamma correction processing is performed on the construction ground grayscale image with overexposure or underexposure to obtain a construction ground contrast enhancement image. The boundary identification layer is used to identify the foundation pit boundary pixel in the construction ground contrast enhancement image and extract the pixel coordinates of the foundation pit boundary pixel. As an embodiment of the present application, the histogram equalization method is the CLAHE (Contrast Limited Adaptive Histogram Equalization) method. The contrast calculation method of the construction ground grayscale image is to calculate the standard deviation and mean value of the pixel grayscale value of the construction ground grayscale image. If the standard deviation is lower than the empirically set preset standard deviation threshold (for example, 25), the construction ground grayscale image is low-contrast. If the mean value is higher than the empirically set preset mean value threshold (for example, 40), the construction ground grayscale image is overexposed. If the mean value is lower than the empirically set preset mean value threshold, the construction ground grayscale image is underexposed. Specifically, for the overexposed construction ground grayscale image, the gamma correction coefficient is set to be greater than 1. For the underexposed construction ground grayscale image, the gamma correction coefficient is set to be less than 1. It should be noted that the contrast enhancement layer can accurately determine whether the image is low-contrast, overexposed, or underexposed by calculating the mean value and standard deviation of the pixel grayscale value of the construction ground grayscale image. CLAHE histogram equalization or gamma correction method is used for processing according to different situations. The gamma coefficient of the overexposed image is set to be greater than 1, and the gamma coefficient of the underexposed image is set to be less than 1, thereby effectively enhancing the texture and details of the construction ground and improving the visibility of the foundation pit edge.

[0023] The boundary recognition layer adopts an improved YOLOv8 model structure, and the improvement manner is: introducing a multi-scale feature fusion structure into the backbone network to enhance the capture ability of the edge and slender structure of the foundation pit; introducing a spatial attention mechanism in the detection head part to improve the positioning accuracy of the boundary points in the weak texture area; adding the foundation pit edge and auxiliary marker dataset in the construction scene in the training stage to improve the robustness of the boundary recognition model under complex light and noise conditions; As a preferred embodiment of the present application, referring to Figure 2 As shown in the figure, the foundation pit boundary pixel recognition process of the boundary recognition layer is: S101: The backbone network in the boundary recognition layer performs multi-scale feature extraction on the construction ground contrast enhanced image, and captures image features at low, medium and high resolutions respectively, wherein the image features at low resolution are used to capture the overall foundation pit contour, and the image features at medium and high resolutions are used to capture the slender boundary lines and auxiliary marker details; S102: The image features of different scales are fused, and the response ability to key boundaries and weak texture areas is enhanced through the spatial attention mechanism; S103: The detection head generates a spatial attention map on the fused feature map, highlights the possible foundation pit boundary pixel area, and generates a boundary probability map on the fused feature map, wherein the boundary probability map is composed of the probability of each pixel being a boundary; S104: The spatial attention map and the boundary probability map are weighted and fused to obtain a boundary weighted probability map; S105: The probability of each pixel being a foundation pit boundary pixel is extracted from the boundary weighted probability map, and the pixels with a probability higher than a preset probability threshold (such as 0.6) are marked as foundation pit boundary pixels; S106: All foundation pit boundary pixels in the construction ground contrast enhanced image are subjected to connected component analysis and boundary smoothing processing to obtain the final foundation pit boundary pixel recognition result; As an embodiment of the present application, the model structure parameter training process of the boundary recognition layer is: A fixed camera is used to shoot the construction site to obtain high-resolution images under different angles and light conditions. The collected high-resolution images cover the entire foundation pit area and the surrounding auxiliary markers (such as guardrails, warning signs, etc.), and the true labels of each pixel in the high-resolution images are manually labeled to form a training set. If the true label is 1, it means that the pixel is a foundation pit boundary pixel, and if the true label is 0, it means that the pixel is not a foundation pit boundary pixel; Based on the training set, a training loss function of the boundary recognition layer is constructed, and the model structure parameters in the boundary recognition layer are trained using Adam or SGD optimizer. The model structure parameters are updated according to the training loss function result through back propagation; The training loss function of the boundary identification layer is : ; Wherein, represents the probability of the i-th pixel in the training set being predicted by the boundary identification layer as a foundation pit boundary pixel, represents the true label of the i-th pixel in the training set, , represents that the i-th pixel in the training set is a foundation pit boundary pixel, represents that the i-th pixel in the training set is not a foundation pit boundary pixel, and S represents the number of pixels in the training set, The smaller the value of the training loss function, the higher the overlap between the predicted probability and the true label, and the higher the identification accuracy of the foundation pit boundary pixel; It should be noted that the present application improves the YOLOv8 model to realize accurate identification of the foundation pit boundary pixel in the construction ground contrast-enhanced image, significantly improving the automation and precision of the construction site monitoring. Specifically, by introducing a multi-scale feature fusion structure into the backbone network, the model can simultaneously capture the overall foundation pit profile at low resolution and the long and narrow boundary line and auxiliary marker information at medium and high resolution, thereby effectively solving the problem that traditional single-scale feature extraction cannot accurately identify complex edges. The introduction of the spatial attention mechanism enables the model to produce a higher response to key boundaries and weak texture areas in the fused feature map, improving the robustness under conditions of light changes, shadow occlusion and noise interference. By generating a spatial attention map and a boundary probability map through the detection head and weighting fusion, the possible foundation pit boundary pixel area can be accurately highlighted, and high-confidence boundary points can be selected by setting a probability threshold, reducing false positives and omissions. Connected component analysis and boundary smoothing further enhance the continuity and realism of the boundary, providing reliable input for subsequent three-dimensional reconstruction, depth measurement and slope stability evaluation. The boundary identification model not only improves the identification accuracy and reliability of the foundation pit boundary pixel, but also reduces the dependence on manual inspection, enabling rapid and accurate acquisition of foundation pit boundary information in complex construction environments, providing a solid data foundation for construction anomaly detection, construction safety warning and management decision-making, thereby significantly improving the intelligent level and construction safety assurance capability of highway construction monitoring; The boundary identification model is used to identify the boundary of the construction ground image, and the pixel coordinates of the identified foundation pit boundary pixel are obtained. The pixel coordinates of the foundation pit boundary pixel are used to map the foundation pit boundary pixel to the highway construction site.

[0024] The pixel coordinates of the foundation pit boundary pixel are used to map the foundation pit boundary pixel to the highway construction site, and the area surrounded by the mapping position of the foundation pit boundary pixel is taken as the foundation pit, comprising: The depth value at the pixel coordinate of the foundation pit boundary pixel is extracted (directly collected by the depth camera), combined with the depth value, and based on an intrinsic matrix of the depth camera, the pixel coordinate of the foundation pit boundary pixel is mapped to a camera coordinate system to obtain three-dimensional camera coordinates of the pixel coordinate of the foundation pit boundary pixel in the camera coordinate system; Specifically, the pixel coordinate The mapping formula for mapping to the camera coordinate system is: ; ; wherein, represents the pixel coordinate mapped to the three-dimensional camera coordinates in the camera coordinate system, represents the depth value of the pixel coordinate , T represents transposition, represents the intrinsic matrix of the depth camera, T represents transposition, represents the inverse matrix of the intrinsic matrix ; specifically, represents the horizontal direction focal length of the depth camera, represents the vertical direction focal length of the depth camera, represents the principal point coordinate of the depth camera; The three-dimensional camera coordinates are converted to a unified highway coordinate system by using an extrinsic rotation matrix and an extrinsic translation vector of the depth camera, to obtain three-dimensional highway coordinates of the three-dimensional camera coordinates in the highway coordinate system; specifically, the conversion to the unified highway coordinate system enables the foundation pit boundary pixels photographed by the depth cameras deployed at different fixed positions to be mapped to the unified highway coordinate system. The conversion formula for converting the three-dimensional camera coordinates to the unified highway coordinate system is: ; wherein, represents the three-dimensional camera coordinates converted to the three-dimensional highway coordinates in the unified highway coordinate system, represents the 3x3 extrinsic rotation matrix of the depth camera, represents the 3x1 extrinsic translation vector of the depth camera; The three-dimensional highway coordinates are connected in the highway coordinate system to form a closed polygon, and the closed polygon is the area surrounded by the mapping positions of the foundation pit boundary pixels.

[0025] Optionally, the three-dimensional highway coordinates can be connected as a closed polygon by using a distance restriction-based connection method, connected domain closure, or curve fitting method, wherein in the distance restriction-based connection method, if the Euclidean distance between the three-dimensional highway coordinates is lower than a preset distance threshold (for example, 0.4 meters), the two three-dimensional highway coordinates are connected.

[0026] S2: controlling the laser radar scanner deployed on the mobile inspection vehicle to collect auxiliary point cloud data of the foundation pit, and controlling the binocular camera deployed on the mobile inspection vehicle to collect a left view and a right view of the foundation pit.

[0027] controlling the laser radar scanner deployed on the mobile inspection vehicle to collect auxiliary point cloud data of the foundation pit, and controlling the binocular camera deployed on the mobile inspection vehicle to collect a left view and a right view of the foundation pit, comprises: deploying the mobile inspection vehicle at appropriate positions around the foundation pits at the highway construction site, to ensure that the mobile inspection vehicle can cover each foundation pit on the inspection path; The mobile inspection vehicle is equipped with a laser radar scanner and a binocular camera, the binocular camera includes a left view camera and a right view camera, respectively used for collecting a left view and a right view; the horizontal distance of the binocular camera has been calibrated to ensure that it can collect image pairs that can be used for stereo matching; the laser radar scanner and the binocular camera are time-synchronized to ensure the consistency of the data collected at the same time in space and time; controlling the mobile inspection vehicle to move along the periphery of the foundation pit, so that the laser radar scanner performs 360° rotary scanning of the foundation pit at a preset scanning frequency (such as 10 Hz), and collects three-dimensional point clouds of the foundation pit and the periphery of the foundation pit to form a three-dimensional point cloud set as auxiliary point cloud data of the foundation pit, the three-dimensional point cloud being in the form of three-dimensional coordinates; Specifically, the collected three-dimensional point cloud is mapped to the highway coordinate system, the three-dimensional point cloud is replaced by the mapped coordinates, and a three-dimensional point cloud set is constructed; controlling the mobile inspection vehicle to simultaneously start the shooting of the left view camera and the right view camera, and collecting a left view and a right view covering the entire panorama of the foundation pit.

[0028] S3: calculating the parallax information of the foundation pit based on the left view and the right view, and combining the auxiliary point cloud data to estimate the depth and width of the foundation pit using a stereo matching algorithm, to obtain the depth and width of the foundation pit.

[0029] calculating the parallax information of the foundation pit based on the left view and the right view, comprises: graying the left view and the right view to obtain a grayed left view and a grayed right view; setting the parallax search range as ; optionally, setting Q as 5 (unit: pixel); ; wherein, represents the parallax value of the pixel coordinate u in the grayed left view under the parallax q, represents the pixel coordinate set of the grayed left view, This represents the left view in grayscale centered at pixel coordinate u. Pixel region, v represents pixel region Any pixel coordinate in the array, This represents the grayscale value at pixel coordinate v in the grayscale left view. This represents the pixel coordinates after the pixel coordinate v has been translated q pixels horizontally. Represents the pixel coordinates in the grayscale right view The grayscale value at that location.

[0030] Combining auxiliary point cloud data, a stereo matching algorithm is used to estimate the depth and width of the foundation pit, obtaining the depth and width of the foundation pit, including: The point cloud consistency matching cost between any pixel coordinate in the grayscale left view and the auxiliary point cloud data is calculated. The point cloud consistency matching cost is the minimum coordinate difference between the pixel coordinate and the three-dimensional point cloud projection result of all three-dimensional point clouds in the auxiliary point cloud data. The larger the coordinate difference, the higher the point cloud consistency matching cost. The three-dimensional point cloud projection result is the projection of the three-dimensional point cloud onto the pixel coordinate system where the pixel coordinate is located in the grayscale left view. The point cloud consistency matching cost between pixel coordinates in the grayscale left view and auxiliary point cloud data, as well as the disparity value of pixel coordinates under different disparities, are added together and used as the initial matching cost of pixel coordinates in the grayscale left view under different disparities. Specifically, the initial matching cost of pixel coordinate u in the grayscale left view under disparity q is: : ; in, This represents the point cloud consistency matching cost for pixel coordinate u in the grayscale left view. This indicates the control weight, which is set based on experience. It is 0.5; The initial matching cost of pixel coordinates in the grayscale left view under different disparities is accumulated in different directions, and the cost accumulation results in multiple directions are summed as the path accumulation cost of pixel coordinates under different disparities. Specifically, the cumulative path cost calculation method for pixel coordinate u in the grayscale left view under disparity q is as follows: ; ; in, This represents the pixel coordinate u in the grayscale left view, along with the disparity q and direction. The cumulative result of the cost, in which direction The value range includes 8 directions: up, down, left, right, and the 4 diagonal directions. pixel coordinate u along direction pixel coordinate u after translation of one pixel, pixel coordinate u in the left view after disparity q and direction accumulation result of the cost, pixel coordinate u in the left view after disparity q+1 and direction accumulation result of the cost, pixel coordinate u in the left view after disparity q-1 and direction accumulation result of the cost; path accumulation cost of pixel coordinate u in the left view after disparity change penalty constant, set to 2; select the minimum value among , , ; select the disparity that minimizes the path accumulation cost of the pixel coordinate u in the left view, generate the disparity depth of the pixel coordinate u in the left view based on the selected disparity, the camera baseline and the focal length, take the disparity depth as the depth value of the pixel coordinate u in the left view, map the pixel coordinate u in the left view to the highway coordinate system in the manner described in step S1, obtain the mapped three-dimensional highway coordinate, take the coordinate value of the mapped three-dimensional highway coordinate on the Z-axis as the depth of the pixel coordinate u in the left view at the foundation pit position; Specifically, the generation formula of the disparity depth is: ; wherein, deep represents the disparity depth, B represents the camera baseline, f represents the focal length of the camera, represents the selected disparity; select the mapped three-dimensional highway coordinate of all pixel coordinates in the left view, project the mapped three-dimensional highway coordinate into a planar coordinate system in the highway coordinate system, obtain the projection coordinate of each mapped three-dimensional working coordinate, calculate the Euclidean distance between any two projection coordinates, and select the maximum Euclidean distance as the width of the foundation pit; specifically, the highway coordinate system is a three-dimensional coordinate system, two axes in the horizontal direction in the highway coordinate system are extracted, and the formed planar coordinate system is the planar coordinate system in the highway coordinate system; select the maximum value of the depth of all pixel coordinates in the left view at the foundation pit position as the depth of the foundation pit.

[0031] It should be noted that the present application can effectively make up for the deficiency of pure pixel grayscale or disparity matching by introducing point cloud consistency matching cost between the grayed left view pixel coordinates and the auxiliary point cloud data. The traditional disparity matching is prone to mismatch in sparse texture, light change or noisy construction scene, while the point cloud projection provides additional three-dimensional geometric constraints, making the matching cost calculation more reliable, thereby improving the accuracy of the initial matching cost.

[0032] In the path cost accumulation stage, the present application adopts a multi-direction (up, down, left, right and four diagonal lines) cost accumulation method, and combines with a disparity change penalty mechanism to effectively suppress the continuous jump of disparity. Compared with the traditional single direction cost accumulation, the present application can reduce the false disparity distribution caused by noise or occlusion. In the depth generation link, an accurate geometric model is introduced based on the camera baseline and focal length to ensure the physical meaning consistency from disparity estimation to depth value, and further to complete the conversion from pixel to actual three-dimensional space by mapping to the highway coordinate system, thereby ensuring that the depth of the foundation pit in the construction scene can intuitively reflect the real topographic depression. Specifically, the method proposed in the present application not only improves the automation level of foundation pit identification and measurement, but also provides reliable basic data support for subsequent safety risk value calculation and secondary inspection decision, thereby improving the intelligibility and safety of the highway construction site inspection.

[0033] S4: dividing the highway construction site into discrete grid regions, generating a safety risk value of the discrete grid based on the foundation pit information of the foundation pit in the discrete grid, and performing secondary inspection on the discrete grid with a safety risk value higher than the allowable risk, wherein the foundation pit information includes the area, depth and width of the foundation pit.

[0034] projecting the highway construction site into a planar coordinate system in the highway coordinate system; specifically, the highway coordinate system is a three-dimensional coordinate system, and two axes in the horizontal direction of the highway coordinate system are extracted, and the formed planar coordinate system is the planar coordinate system in the highway coordinate system; dividing the projected highway construction site into a plurality of square discrete grids; Optionally, the side length Len of the square discrete grid satisfies wherein is the side length range of the set square discrete grid; The side length of the discrete grid can be adjusted arbitrarily within the side length range, so that the discrete grid can cover the projected highway construction site, and for the foundation pit spanning the discrete grid, the side length of the discrete grid is adjusted so that the same foundation pit is completely contained in a single discrete grid, ensuring that each foundation pit belongs to only one discrete grid; The safety risk value of the discrete grid is generated based on the pit information of the pits in the discrete grid.

[0035] Optionally, the allowable risk is set according to quantiles of the safety risk values of all discrete grids in engineering specifications or expert experience or history, and the allowable risk is set according to quantiles of the safety risk values of all discrete grids in history in the following manner: the safety risk value of the quantile of 10% is selected as the allowable risk by obtaining the safety risk values of all discrete grids in the same period of highway construction in history and performing descending sorting, wherein the safety risk value of the quantile of 10% is the safety risk value only lower than the obtained safety risk value of 10%.

[0036] The safety risk value calculation formula of the discrete grid is as follows: wherein R represents the safety risk value of the discrete grid, Len represents the side length of the discrete grid, H represents the number of pits in the discrete grid, represents the area of the hth pit in the discrete grid, represents the depth of the hth pit in the discrete grid, represents the width of the hth pit in the discrete grid, represents the maximum depth of all pits in the highway construction site, the normalized values of the area , the depth and the width in turn, respectively, and represents the volume proportion factor of the hth pit; optionally, the area calculation manner of the pit in the discrete network is as follows: the projection coordinates of the mapped three-dimensional working coordinates are marked in the planar coordinate system in the highway coordinate system, the marked area is taken as the pit plane under the planar coordinate system, and the area of the minimum circumscribed polygon surrounding the pit plane is calculated in the planar coordinate system. Specifically, reflects the overall damage degree of the hth pit to the surface of the discrete grid, comprehensively evaluates the safety risk in combination with the depth and width of the hth pit, introduces a nonlinear amplification mechanism to make the pit risk rise exponentially, reflects the overall weakening effect of the hth pit on the local foundation stability, and is suitable for identifying small high-risk pits with small area but deep depth; both represent effect coefficients, represents a nonlinear amplification coefficient; based on experience are set to 0.3, 0.4, 0.3 and 1.2 in turn.

[0037] ​​It should be appreciated that the above-described embodiments are merely illustrative, and the patent application scope is not limited by the structure.

[0038] It should be noted that the above-mentioned embodiment numbers of the present application are only for description, and do not represent the advantages and disadvantages of the embodiments. And the term "include", "contain" or any other variant thereof in this paper is intended to cover non-exclusive inclusion, so that the process, device, article or method including a series of elements not only includes those elements, but also includes other elements not explicitly listed, or includes elements inherent to such process, device, article or method. Without more limitations, the element defined by the sentence "including a" does not exclude the existence of other identical elements in the process, device, article or method including the element.

[0039] From the above description of the embodiments, those skilled in the art can clearly understand that the above-mentioned embodiment method can be realized by software and the necessary general hardware platform, of course, it can also be realized by hardware, but in many cases the former is a better embodiment. Based on such understanding, the technical solutions of the present application can be embodied in the form of a software product, which is stored in a storage medium (such as ROM / RAM, magnetic disk, optical disk) as described above, and includes a plurality of instructions for making a terminal device (which can be a mobile phone, computer, server, or network device, etc.) execute the method described in each embodiment of the present application.

[0040] The above is only the preferred embodiment of the present application, and does not limit the patent scope of the present application, and any equivalent structure or equivalent process transformation made by using the content of the present application specification and drawings, or directly or indirectly applied to other related technical fields, are also included in the patent protection scope of the present application.

Claims

1. An expressway construction inspection optimization method based on artificial intelligence, characterized in that, The method comprises: S1: collecting the construction ground image of the highway construction site by using the fixed position depth camera, identifying the foundation pit boundary pixels in the construction ground image by using the boundary identification model, mapping the foundation pit boundary pixels to the highway construction site, and taking the area surrounded by the mapping position of the foundation pit boundary pixels as the foundation pit; S2: controlling the laser radar scanner deployed on the mobile inspection vehicle to collect auxiliary point cloud data of the foundation pit, and controlling the binocular camera deployed on the mobile inspection vehicle to collect the left view and the right view of the foundation pit; S3: calculating the parallax information of the foundation pit based on the left view and the right view, and combining the auxiliary point cloud data, using a stereo matching algorithm to estimate the depth and width of the foundation pit, and obtaining the depth and width of the foundation pit; S4: dividing the highway construction site into discrete grid areas, generating safety risk values of the discrete grid based on the foundation pit information of the foundation pit in the discrete grid, and performing secondary inspection on the discrete grid with a safety risk value higher than the allowable risk, wherein the foundation pit information includes the area, depth and width of the foundation pit.

2. The highway construction inspection optimization method based on artificial intelligence according to claim 1, wherein, The boundary identification model comprises an input layer, a contrast enhancement layer and a boundary identification layer. The input layer is used to receive the construction ground image and perform grayscale processing to obtain a construction ground grayscale image. The contrast enhancement layer is used to calculate the contrast of the construction ground grayscale image, and the histogram equalization method is used for contrast enhancement processing on the construction ground grayscale image with low contrast. The gamma correction processing is performed on the overexposed or underexposed construction ground grayscale image to obtain a construction ground contrast enhancement image. The boundary identification layer is used to identify the foundation pit boundary pixels in the construction ground contrast enhancement image, obtain the pixel coordinates of the foundation pit boundary pixels, and extract the pixel coordinates of the foundation pit boundary pixels. The boundary identification layer uses an improved YOLOv8 model structure, and the improvement method is: introducing a multi-scale feature fusion structure into the backbone network; introducing a spatial attention mechanism into the detection head part. The boundary identification model is used to identify the boundaries of the construction ground image, and the pixel coordinates of the foundation pit boundary pixels are obtained. The pixel coordinates of the foundation pit boundary pixels are used to map the foundation pit boundary pixels to the highway construction site. The pixel coordinates of the foundation pit boundary pixels are extracted, and the depth values at the pixel coordinates are obtained. Based on the depth values and the intrinsic matrix of the depth camera, the pixel coordinates are mapped to the camera coordinate system to obtain the three-dimensional camera coordinates of the foundation pit boundary pixels in the camera coordinate system. The three-dimensional camera coordinates are converted to a unified highway coordinate system by using the extrinsic rotation matrix and the extrinsic translation vector of the depth camera to obtain the three-dimensional highway coordinates of the three-dimensional camera coordinates in the highway coordinate system. ​ 3. The highway construction inspection optimization method based on artificial intelligence according to claim 2, characterized in that, ​ ​ ​ The three-dimensional road coordinates are connected in the expressway coordinate system to form a closed polygon, and the closed polygon is an area surrounded by the pixel mapping position of the foundation pit boundary.

4. The highway construction inspection optimization method based on artificial intelligence according to claim 1, wherein, The laser radar scanner deployed on the mobile inspection vehicle collects auxiliary point cloud data of the foundation pit, and the binocular camera deployed on the mobile inspection vehicle collects left and right views of the foundation pit, including: In the expressway construction site, the mobile inspection vehicle is deployed at a suitable position around the foundation pit to ensure that the mobile inspection vehicle can cover each foundation pit on the inspection path; The mobile inspection vehicle is equipped with a laser radar scanner and a binocular camera, and the binocular camera includes a left view camera and a right view camera, which are used to collect left and right views, respectively. The mobile inspection vehicle is controlled to move around the foundation pit, and the laser radar scanner is controlled to rotate around the foundation pit at a preset scanning frequency to collect three-dimensional point clouds of the foundation pit and the surrounding area, forming a three-dimensional point cloud set as auxiliary point cloud data of the foundation pit, wherein the three-dimensional point cloud is in the form of three-dimensional coordinates. The mobile inspection vehicle is controlled to start the shooting of the left and right view cameras simultaneously to collect left and right views covering the entire panorama of the foundation pit.

5. The highway construction inspection optimization method based on artificial intelligence according to claim 4, characterized in that, Based on the left and right views, the disparity information of the foundation pit is calculated, including: The left and right views are subjected to grayscale processing to obtain grayscale left and right views. The disparity search range is set to ; The disparity values of any pixel coordinates in the grayscale left view under different disparities are calculated as the disparity information of the foundation pit. ； wherein, represents a disparity value of a pixel coordinate u in the left view in the gray scale at a disparity q, represents a set of pixel coordinates of the left view in the gray scale, represents a pixel region centered at a pixel coordinate u in the left view in the gray scale, v represents an arbitrary pixel coordinate in the pixel region , represents a gray scale value at a pixel coordinate v in the left view in the gray scale, represents a pixel coordinate after the pixel coordinate v is shifted by q pixels in the horizontal direction, represents a gray scale value at a pixel coordinate in the right view in the gray scale; Auxiliary point cloud data of the foundation pit is obtained.

6. The highway construction inspection optimization method based on artificial intelligence according to claim 5, characterized in that, Combined with the auxiliary point cloud data, a stereo matching algorithm is used to estimate the depth and width of the foundation pit to obtain the depth and width of the foundation pit, including: The point cloud consistent matching cost between any pixel coordinates in the grayscale left view and the auxiliary point cloud data is calculated, wherein the point cloud consistent matching cost is the smallest coordinate difference between the pixel coordinates and the three-dimensional point cloud projection results of all three-dimensional point clouds in the auxiliary point cloud data, and the three-dimensional point cloud projection result is the projection of the three-dimensional point cloud to the pixel coordinate system in which the pixel coordinates are located. The point cloud consistent matching cost between the pixel coordinates in the grayscale left view and the auxiliary point cloud data and the disparity values of the pixel coordinates under different disparities are added together as the initial matching cost of the pixel coordinates under different disparities. The initial matching cost of the pixel coordinates under different disparities is accumulated in different directions, and the sum of the cost accumulation results in multiple directions is taken as the path accumulation cost of the pixel coordinates under different disparities. The disparity that minimizes the path accumulation cost of the pixel coordinates in the grayscale left view is selected, and the disparity depth of the pixel coordinates in the grayscale left view is generated based on the selected disparity, the camera baseline, and the focal length. The disparity depth is taken as the depth value of the pixel coordinates in the grayscale left view. The pixel coordinates in the grayscale left view are mapped to the expressway coordinate system in the manner described in step S1, and the mapped three-dimensional road coordinates are obtained. The coordinate value of the mapped three-dimensional road coordinates on the Z-axis is taken as the depth of the pixel coordinates in the grayscale left view at the foundation pit position. The mapped three-dimensional road coordinates of all pixel coordinates in the grayed left view are selected, and the mapped three-dimensional road coordinates are projected into a plane coordinate system in the highway coordinate system to obtain the projection coordinates of each mapped three-dimensional working coordinate, the Euclidean distance between any two projection coordinates is calculated, and the maximum Euclidean distance is selected as the width of the foundation pit. The maximum value of the depth of all pixel coordinates in the grayed left view at the position of the foundation pit is selected as the depth of the foundation pit.

7. The highway construction inspection optimization method based on artificial intelligence according to claim 1, characterized in that, Step S4 comprises: projecting the highway construction site into a plane coordinate system in the highway coordinate system; dividing the projected highway construction site into a plurality of square discrete grids; generating a safety risk value of the discrete grid based on the foundation pit information of the foundation pit in the discrete grid.

8. The highway construction inspection optimization method based on artificial intelligence according to claim 7, characterized in that, The safety risk value calculation formula of the discrete grid is: ； wherein R represents the safety risk value of the discrete grid, Len represents the side length of the discrete grid, H represents the number of foundation pits within the discrete grid, represents the area of the hth foundation pit within the discrete grid, represents the depth of the hth foundation pit within the discrete grid, represents the width of the hth foundation pit within the discrete grid, represents the maximum depth of all foundation pits in the expressway construction site, the normalized values of the area , the depth and the width in turn, respectively, represents the volume proportion factor of the hth foundation pit; all represent the effect coefficient, represents the nonlinear amplification coefficient.

Images