Concrete pit measuring and reporting method and system based on scale marker

Abstract

This invention discloses a method and system for measuring and reporting concrete pits based on scale markers, belonging to the field of geometric measurement technology. The method involves deploying visual scale markers with known true side lengths near the pit and acquiring multi-view images. After quality screening and corner point extraction, a scene point cloud is generated through visual 3D reconstruction. Based on the 3D coordinates of the marker corner points, a scale coefficient is obtained through robust multi-view statistics to restore the true physical scale of the point cloud. Further point cloud preprocessing, reference plane fitting, and pit region segmentation automatically calculate the pit's projected area, depth, volume, and contour parameters. Combined with quality evaluation, the measurement results are automatically reported to a remote management platform through the task information carried by the markers. This invention eliminates the need for dedicated high-precision scanning equipment, enabling low-cost 3D measurement and end-to-end information reporting of true-scale pits, and is suitable for large-scale on-site inspection scenarios.

Description

Technical Field

[0001] This invention relates to the field of geometric measurement technology, and in particular to a method and system for non-contact measurement of the depth, area and contour of depressions on the surface of concrete components using visual scale markers, and for automatically filling in the inspection records. Background Technology

[0002] Concrete components are susceptible to surface spalling and localized pitting damage during service due to factors such as vehicle loads, freeze-thaw cycles, chloride erosion, and external impacts. It is necessary to record and assess the location, area, depth, and repair work required for these damages. Current engineering inspections rely heavily on manual visual inspection, measurement, photography, and form filling, which are significantly affected by personnel experience, shooting angles, and site conditions, making it difficult to consistently obtain three-dimensional geometric information such as the depth and volume of pits. While laser scanning and structured light scanning equipment offer higher precision, they are costly and complex to deploy, hindering large-scale inspections.

[0003] Vision-based 3D reconstruction models can generate dense point clouds or depth maps from multi-view images, providing a technical path for low-cost 3D measurement. However, the point clouds obtained from vision reconstruction are usually in the model's own reconstruction coordinate system, lacking a scale correspondence with the real world. Without scale correction, directly calculating the area, depth, and volume of pits from the reconstructed point clouds will result in significant scale deviations. Existing calibration rulers, calibration boards, and other reference objects only serve a scale control function and are difficult to bind simultaneously with the object number, form entry, and remote platform. Although QR codes can carry information, existing solutions mostly use them as information entry points and rarely use them as composite identifiers for scale correction, task binding, and automatic reporting triggering.

[0004] Furthermore, the existing inspection process involves a disconnect between geometric information extraction and test record filling. The area and depth of the pits obtained from point cloud computing still require manual input into questionnaires, forms, or management platforms, which is prone to problems such as transcription errors, unit conversion errors, mismatches between photos and records, and untimely remote management. Therefore, there is an urgent need for a low-cost method for measuring concrete pits that allows for the recovery of true-scale data and the automatic extraction and filling of three-dimensional geometric information. Summary of the Invention

[0005] To address the technical problems mentioned in the background art, such as the lack of true scale in visual reconstruction results and the disconnect between pit measurement and detection record filling, this invention provides a method and system for measuring and filling concrete pits based on scale markers. By deploying visual scale markers near the pits, which combine scale correction, task binding, and filling entry functions, the entire process of pit three-dimensional geometric information is realized, from multi-view image acquisition, true scale restoration, automatic measurement to automatic filling and remote management.

[0006] To achieve the above objectives, the present invention provides the following technical solution: In a first aspect, the present invention provides a method for measuring and reporting concrete pits based on dimensional markers, comprising the following steps: S1. Visual scale markers are set up on the flat surface near the concrete pit to be tested. The visual scale markers have known true side lengths and carry detection task information. Multi-view image sequences containing the pit, the surrounding flat concrete surface and the visual scale markers are acquired from multiple perspectives by an imaging device. S2. Perform quality screening on the images in the multi-view image sequence, identify the visual scale markers in the screened images, and extract the corner pixel coordinates of the visual scale markers; S3. Input the selected multi-view images into the visual 3D reconstruction model to generate a scene point cloud containing pits, flat concrete surfaces and the visual scale markers, and establish the correspondence between image pixels and 3D points. S4. Obtain the corresponding three-dimensional point coordinates based on the corner pixel coordinates, calculate the side length of the visual scale marker in the reconstructed coordinate system, use the known real side length as a reference, combine the side length to obtain the scale coefficient through robust statistics, and use the scale coefficient to convert the reconstructed point cloud into a real scale point cloud. S5. Preprocess the real-scale point cloud, fit the local reference plane with the point cloud of the flat concrete surface around the pit, calculate the signed distance of each point in the point cloud to the local reference plane, and extract the continuous area below the local reference plane as the pit area. S6. Based on the point cloud of the pit region and the local reference plane, calculate the projected area, depth index, volume and contour parameters of the pit; S7. Read the detection task information carried by the visual scale marker, automatically fill in the projected area, depth index, volume and contour parameters of the pit and the acquired image into the online form corresponding to the detection task information, and synchronize it to the management platform for remote viewing and traceability.

[0007] Furthermore, the visual scale marker is a visual target, which can be any one of a QR code, a square sign containing a QR code, a composite target with a pattern of known geometric dimensions, an ArUco logo, or an AprilTag logo; the detection task information includes the detection object number, component position, actual side length of the marker, form address, server address, detection task number, and timestamp.

[0008] The beneficial effect of the above-mentioned further solutions is that the identification device can simultaneously perform the functions of scale calibration, object binding, and remote reporting entry, thus achieving multiple uses for one device.

[0009] Optionally, the visual scale marker can also be other targets with identifiable patterns of known geometric dimensions, such as checkerboard targets, dot array targets, or reflective markers with known side lengths. As long as the target can be stably identified by the imaging device, provides extractable corner points or feature points, and its true size is known, it falls within the protection scope of this invention.

[0010] Furthermore, in step S2, the quality screening includes sharpness evaluation, exposure evaluation, marker integrity evaluation, and image quantity evaluation; the sharpness evaluation is achieved through any one of Laplacian variance, edge strength, and depth reconstruction confidence; the marker integrity evaluation is achieved through decoding results, corner number, and corner sorting consistency; for images that fail to identify markers, their role in scale estimation is removed, but their right to participate in 3D reconstruction is retained.

[0011] The beneficial effect of the above-mentioned further scheme is that it balances the accuracy of scale estimation with the utilization rate of reconstructed images, and avoids image waste.

[0012] Furthermore, in step S3, the visual 3D reconstruction model outputs a depth map, camera intrinsic parameters, camera pose, and world point map. The world point map is a 3D coordinate array that corresponds one-to-one with the image pixel coordinates, used to establish the correspondence between image pixels and 3D points.

[0013] The beneficial effect of the above-mentioned further scheme is that it provides a stable source of three-dimensional points for scale correction.

[0014] Optionally, the visual 3D reconstruction model can also output pixel-level depth confidence for each image, which can be used to weight or remove low-confidence points during subsequent point cloud fusion and reference plane fitting, thereby further improving the reliability of pit measurement.

[0015] Further, in step S4, obtaining the scaling coefficients through robust statistics includes: For each usable image, the reconstructed length of each side is calculated based on the three-dimensional coordinates of each corner point of the visual scale marker, and the scale candidate value of the image is obtained by combining the known true side length. The median, interquartile range, and median absolute deviation are used to remove anomalies from the scale candidate values ​​of multiple images. After removing abnormal viewpoints, the median of the remaining scale candidate values ​​is used as the final scale coefficient. At the same time, the standard deviation, coefficient of variation, number of valid images, and number of abnormal images of the scale estimation are output. The reconstruction point cloud is converted into a true scale point cloud using the scale coefficient, which includes multiplying length indicators by the scale coefficient, area indicators by the square of the scale coefficient, and volume indicators by the cube of the scale coefficient.

[0016] The beneficial effect of the above-mentioned further scheme is that it eliminates abnormal scale values ​​caused by QR code occlusion, corner point false detection and excessive viewing angle through multi-view robust statistics, thereby improving the reliability of scale recovery.

[0017] Optionally, when the difference between the reconstructed lengths of each side of the marker exceeds a preset threshold from a certain perspective, the perspective is determined to be an abnormal perspective. The preset threshold can be set according to the ratio of the actual side length of the marker. For example, when the ratio of the maximum value to the minimum value among the reconstructed lengths of each side is greater than 1.2, it is determined to be an abnormal perspective.

[0018] Further, in step S5, the preprocessing includes statistical outlier filtering, radius outlier filtering, voxel downsampling, spatial range clipping, and voxel connected component analysis; the local reference plane is obtained by fitting any one of RANSAC, least squares fitting, robust plane fitting, and region growing methods, and is used to characterize the original concrete surface before the pit occurs; the extraction of continuous regions below the local reference plane as pit regions specifically involves: screening candidate regions using depth thresholds, connectivity constraints, and region morphology constraints, and removing isolated noise points through radius filtering, retention of the largest connected component, and boundary smoothing.

[0019] The beneficial effects of the above-mentioned further solutions are: reducing the impact of reconstruction noise, background points, and non-target component points on geometric measurements, and improving the accuracy of pit segmentation.

[0020] Further, in step S6, the projected area is calculated using any one of the following methods: regular grid method, AlphaShape, convex hull method, and boundary polygon method; the depth index includes maximum depth, average depth, and root mean square depth; the volume is calculated using the height map integral method: the projected area of ​​the pit is divided into a two-dimensional grid, and the mean, median, and quantile of the pit depth in each grid are used as the representative depth of that grid, and the volume is the sum of the products of the area of ​​each occupied grid and the representative depth; the contour parameters include principal direction length, width, aspect ratio, perimeter, boundary contour, and depth heatmap.

[0021] The beneficial effects of the above-mentioned further solutions are: outputting multi-dimensional three-dimensional geometric indicators covering area, depth, volume and contour, providing data support for maintenance design, material usage estimation and detailed damage assessment.

[0022] Furthermore, the process also includes: prior to step S7, evaluating the quality of scale estimation, point cloud reconstruction, reference plane fitting, and pit segmentation. The quality evaluation indicators include: scale estimation quality, characterized by the standard deviation, coefficient of variation, and number of effective views of the multi-view scale coefficients; point cloud reconstruction quality, characterized by point cloud density, integrity, and reconstruction confidence; reference plane fitting quality, characterized by the proportion of points in the plane, fitting residuals, and normal stability; and pit segmentation quality, characterized by the area of ​​connected components, continuity of depth distribution, and proportion of outliers. If the quality meets the standards, the process proceeds to the automatic reporting process of step S7; if the quality does not meet the standards, a prompt for re-collection is made. The measurement results after manual confirmation proceed to the automatic reporting process of step S7.

[0023] The beneficial effects of the above-mentioned further solutions are: to comprehensively evaluate the reliability of measurements from four dimensions: scale, reconstruction, fitting, and segmentation, and to avoid low-quality data from being directly included in engineering records.

[0024] Optionally, the quality evaluation may also include a geometric result integrity evaluation, that is, verifying whether the projected area, depth index, volume and contour parameters of the pit have all been successfully calculated and the values ​​are within a reasonable range; when any index is missing or exceeds the reasonable range, it will also prompt for re-collection or manual confirmation.

[0025] Secondly, this invention also proposes a concrete pit measurement and reporting system based on scale markers, comprising: an image acquisition and marker placement unit, used to place visual scale markers near the concrete pit to be measured, wherein the visual scale markers have known true side lengths and carry detection task information, and guide imaging equipment to acquire multi-view image sequences; an image quality control and marker recognition unit, used to perform quality screening on the acquired images, identify the visual scale markers, and extract the corner pixel coordinates; a visual 3D reconstruction unit, used to generate scene point clouds, depth maps, and the correspondence between pixel points and 3D points based on the multi-view images; and a scale correction unit, used to, based on the 3D coordinates of the marker corners and the known true side lengths, […]. A robust statistical calculation unit with multiple viewpoints calculates the scale coefficient, converting the reconstructed point cloud into a real-scale point cloud. A reference plane extraction and pit segmentation unit preprocesses the real-scale point cloud, fits a local reference plane, and extracts the pit region point cloud. A geometric information calculation unit calculates the projected area, depth, volume, and contour parameters of the pits. A quality evaluation unit evaluates the quality of scale estimation, point cloud reconstruction, reference plane fitting, and pit segmentation, outputting anomaly alerts. An automatic reporting and remote management unit reads the detection task information carried by the visual scale marker, automatically fills the measurement results and supplementary data into the online form corresponding to the detection task information, and synchronizes it to the management platform for remote viewing and traceability.

[0026] Each unit of the system executes the corresponding steps of the method described in the first aspect, supporting the distributed deployment of image acquisition and 3D reconstruction, scale correction, pit segmentation and geometric calculation on mobile terminals, edge computing devices, workstations and cloud servers, and transmitting data through the network to adapt to engineering inspection scenarios of different scales.

[0027] Furthermore, the concrete pit measurement and filling system is deployed on the same computing terminal, or on a mobile terminal, edge computing device, workstation, and cloud server, and transmits data through a network; the computing terminal is any one of a mobile terminal, edge computing device, workstation, and cloud server; the imaging device is any one of a mobile phone, tablet computer, and drone equipped with a camera.

[0028] The beneficial effects of the above-mentioned further solutions are: they support both single-terminal integrated offline measurement and distributed deployment of mobile terminal data collection, cloud computing and remote management, and the deployment method can be flexibly selected according to the on-site network conditions and computing resources.

[0029] Optionally, the system also includes a task management unit, which is used to pre-create inspection tasks on the management platform side and generate inspection task information containing the inspection object number, component location and form address, and encode the inspection task information into the identification component when the visual scale identification component is made, so as to realize the on-site scanning and quick binding of inspection tasks and automatic result backfilling.

[0030] The beneficial effects of this invention are as follows: By using a visual scale marker to simultaneously perform scale calibration, object binding, and remote reporting functions, a complete process is formed, encompassing multi-view image acquisition, 3D reconstruction, true scale restoration, pit geometry measurement, and automatic reporting of inspection records. This achieves low-cost, true-scale 3D measurement of pits and end-to-end information reporting without the need for dedicated high-precision scanning equipment. Using ordinary mobile phone multi-view images as the primary data source, it does not rely on laser scanners, structured light scanners, or dedicated depth cameras, reducing equipment costs and on-site deployment difficulties for 3D measurement of concrete pits, making it suitable for large-scale engineering inspection applications. The true-scale restoration of the reconstructed point cloud is achieved through QR codes or visual scale markers of known dimensions, solving the problem of mismatch between visual 3D reconstruction results and real-world scales, allowing pit area, depth, and volume to be output in engineering units such as millimeters, square millimeters, or cubic centimeters. Integrating the scale calibration, object binding, and form navigation functions of the QR code into a single marker allows it to serve as both a scale control object for 3D measurement and a detection task and remote reporting entry point, improving the consistency and traceability of on-site inspection records. It can not only output the projected area of ​​pits, but also further output three-dimensional geometric information such as maximum depth, average depth, volume, and depth map, providing richer data support than a single area index for maintenance method selection, material usage estimation, and refined damage assessment. Through reference plane fitting, signed distance calculation, and connected component filtering, it automatically separates pits from the surrounding smooth concrete surface, reducing subjective errors caused by manual boundary delineation and readings, and improving the stability of measurement results. The introduction of multi-view scale estimation and robust statistical mechanisms can eliminate scale anomalies caused by QR code occlusion, corner point false detection, excessively oblique viewing angles, or local reconstruction anomalies, thereby improving the reliability of scale correction. Attached Figure Description

[0031] Figure 1 This is a flowchart illustrating the method for measuring and reporting concrete pits based on dimensional markers provided in an embodiment of the present invention. Figure 2 This is a schematic diagram of the concrete pit measurement and reporting system based on scale markers provided in an embodiment of the present invention. Detailed Implementation

[0032] The present invention will be further described in detail below with reference to the accompanying drawings and specific embodiments.

[0033] In this invention, visual scale markers refer to man-made markers with known true geometric dimensions, recognizable by imaging devices, and capable of carrying coded information. In this embodiment, a square QR code is used as the visual scale marker. Other visual targets with known dimensions can also be used as visual scale markers, such as ArUco and AprilTag markers. Both ArUco and AprilTag markers are well-known open-source visual positioning targets: ArUco markers consist of a black border and an internal binary matrix, each marker having a unique ID code, which can be used for camera pose estimation and size calibration; AprilTag markers are binary square targets with similar principles, exhibiting high detection robustness and corner point positioning accuracy. All of the above markers can be generated and decoded using general QR code generation and recognition algorithms, and their true side lengths can be measured and entered into the system before deployment. Therefore, they can all serve as visual scale markers in this invention, replacing square QR codes to achieve scale correction and task information carrying, and all fall within the protection scope of this invention.

[0034] In this invention, robust statistics refers to the method of estimating and removing outliers by using statistical measures such as median, interquartile range, and median absolute deviation, which are insensitive to outliers, when making statistical estimates of multiple sets of observation data containing outliers, so as to avoid the overall deviation of the scaling coefficient caused by individual outliers.

[0035] In this invention, RANSAC is an abbreviation for Random Sample Consensus, a well-known robust fitting algorithm that estimates model parameters through random sampling and interior point counting iterations, and is used to robustly fit a local reference plane in noisy point clouds.

[0036] In this invention, the regular grid method refers to dividing the area containing the planar projection points of the pit into a regular two-dimensional grid, and approximating the projected area of ​​the pit by multiplying the number of grids occupied by the area of ​​a unit grid; AlphaShape is the Alpha shape method, which uses the parameter alpha to control the degree of boundary contraction and generate a compact polygonal boundary from the planar point set, which can refine the concavity of the convex hull of the pit projection point set; the convex hull method is a method to calculate the projected area using the area of ​​the convex hull polygon of the planar point set; the boundary polygon method is a method to directly construct a closed polygon from the boundary points of the pit projection and calculate its area. All of the above methods can obtain the planar projected area of ​​the pit when the pit projection point set is known, and are well-known geometric calculation methods in this field.

[0037] In this embodiment, a square QR code is used as a visual scale marker to perform three-dimensional measurement and information reporting of pit damage on the surface of concrete components.

[0038] like Figure 1 As shown, the specific implementation steps of the method are as follows: Step S1: Visual scale marker placement and multi-view image acquisition.

[0039] Near the pit or spalling area of ​​the concrete to be tested, select a relatively flat concrete area on the same local component surface as the pit, and affix or place a visual scale marker with a known true side length. The side length of the visual scale marker can be pre-entered into the system or encoded in QR code information and read along with the inspection task. The QR code information may include at least one or more of the following: the inspection object number, component location, true side length of the marker, form or server address, inspection task number, and timestamp.

[0040] The data acquisition device can be a regular mobile phone, tablet, drone equipped with a camera, or other device capable of acquiring RGB images. During acquisition, multiple images are taken from different angles around the pit area, ensuring that each or most images simultaneously include the pit, the surrounding smooth concrete surface, and the visually scaled marker. The system can prompt the acquisition personnel to maintain an appropriate overlap, shooting distance, and angle of view, and will filter out or prompt for retaking images that are blurry, overexposed, have incomplete markers, or have excessive repetition of perspective.

[0041] Step S2: Image quality screening, QR code recognition, and corner point extraction.

[0042] The system performs quality screening on the acquired images, which includes at least sharpness evaluation, exposure evaluation, marker integrity evaluation, and image quantity evaluation. Sharpness can be judged by Laplacian variance, edge strength, or depth reconstruction confidence; marker integrity can be judged by QR code decoding results, corner count, and corner sorting consistency.

[0043] For images that pass the quality screening, the system identifies the four corner points of the QR code or visual identifier and obtains the pixel coordinates of the corner points. If a square QR code is used, the corner points can be represented as q1, q2, q3, and q4, and ordered in a clockwise or counterclockwise direction. For images with occluded corner points or failed recognition, the system can remove the image's contribution to scale estimation, but can still decide whether to include it in 3D reconstruction based on image quality.

[0044] Step S3: Visual 3D reconstruction based on multi-view images.

[0045] The system inputs selected multi-view images into a visual 3D reconstruction model to generate scene point clouds, depth maps, camera intrinsics, camera poses, or world point maps. In this implementation, a visual image reconstruction model can be used to obtain the depth information and world point map corresponding to each image. The world point map is a 3D coordinate array that corresponds one-to-one or approximately to the image pixel coordinates. The system can also employ other multi-view geometric, neural 3D reconstruction, or depth estimation models, as long as they can provide data to establish the correspondence between image pixels and 3D points.

[0046] The point cloud output by the visual 3D reconstruction unit is usually located in the reconstruction coordinate system, and its scale is inconsistent with the scale of the real world. Therefore, scale correction is required in subsequent steps using visual scale markers.

[0047] Step S4: Multi-view scale correction based on the three-dimensional coordinates of the corner points of the marker.

[0048] Based on the corner pixel coordinates obtained in step S2, the system combines the back projection results of the world point map and depth map output in step S3 with the camera intrinsic parameters, or samples the corresponding 3D points from the reconstructed point cloud. For each usable image, the system obtains the 3D coordinates P1, P2, P3, and P4 of the four corner points of the QR code in the reconstructed coordinate system, and calculates the reconstructed lengths of the four sides and the diagonal length.

[0049] If the actual side length of the QR code is L, and the reconstructed length of the k-th side in an image is lk, then the scale factor corresponding to that side can be expressed as L / lk. The system can calculate the median, mean, or weighted average of the four sides of the same image to obtain the candidate scale values ​​for that image; then, robust statistics are performed on the candidate scale values ​​obtained from multiple images to obtain the final scale factor.

[0050] In this implementation, the system uses the median and interquartile range or median absolute deviation to remove anomalies from the scale candidate values. When the reconstructed side length of the QR code corner points in an image has excessive differences, the diagonal ratio is abnormal, the corner point confidence is low, or the scale candidate value deviates from the overall statistical range, the system identifies it as an abnormal viewpoint and removes it. The final scale coefficient can be expressed as the median of the scale candidate values ​​after anomaly removal. The system also outputs the standard deviation of the scale estimate, the coefficient of variation, the number of valid images, and the number of abnormal images for quality evaluation.

[0051] Scale correction can be applied to the original point cloud coordinates or to the geometric calculation results. For length-related indicators, the reconstructed units are multiplied by a scale factor to convert them to centimeters or millimeters; for area-related indicators, they are multiplied by the square of the scale factor; and for volume-related indicators, they are multiplied by the cube of the scale factor. For ease of uniform processing, in this implementation, the system first converts the point cloud coordinates to true-scale coordinates before performing reference plane fitting, pit segmentation, and geometric calculations.

[0052] Step S5: Preprocessing of the scaled point cloud, fitting of the reference plane, and segmentation of the pit region.

[0053] The system preprocesses the scale-corrected point cloud, including one or more of the following: statistical outlier filtering, radius outlier filtering, voxel downsampling, color or spatial range cropping, and voxel connectivity analysis. The purpose of preprocessing is to reduce the impact of reconstruction noise, background points, floating points, and non-target component points on subsequent calculations.

[0054] The system fits a local reference plane using points on the smooth concrete surface surrounding the pit. The reference plane can be obtained through RANSAC, least-squares fitting, robust plane fitting, or region growing methods. This reference plane is used to approximately represent the original concrete surface before the pit occurred.

[0055] The system calculates the signed distance from each point in the point cloud to the reference plane and determines the pit depth direction based on the normal direction. For points below the reference plane, the system obtains the corresponding depth value based on the reference plane, or takes the absolute pitting distance based on the normal direction. The system extracts candidate pit regions using depth thresholds, connectivity constraints, and region morphology constraints.

[0056] To avoid misclassifying isolated noise points as pits, the system can perform radius outlier filtering, voxel connected component maximum block retention, and boundary smoothing operations on the pit candidate region. If multiple pit regions exist, the system can select the target pit based on QR code task information, user-specified region, or area / depth sorting, or it can output multiple pits separately with different numbers.

[0057] Step S6: Automatic calculation of pit depth, area, volume, and contour parameters.

[0058] The system establishes a two-dimensional local coordinate system within a reference plane and projects the pit point cloud onto this coordinate system to obtain the planar projection point set of the pit. The system can calculate the projected area of ​​the pit using the regular mesh method, AlphaShape, convex hull method, or boundary polygon method. Based on the depth values ​​from the pit points to the reference plane, the system calculates indicators such as maximum depth, average depth, and root mean square depth. The maximum depth reflects the most severe local depression, while the average depth reflects the overall damage level.

[0059] The system can calculate the volume of a pit based on the heightmap integral method. Specifically, the projected area of ​​the pit is divided into a two-dimensional grid. The mean, median, or preset quantile of the pit point depth within each grid is taken as the representative depth of that grid. The pit volume is approximately the sum of the products of the area occupied by each grid and the representative depth. The system can also output the pit's principal direction length, width, aspect ratio, perimeter, boundary contour, depth heatmap, and 3D point cloud file.

[0060] In the implementation, the system can perform unit conversions according to the requirements of the engineering form, such as converting the projected area from square centimeters to square millimeters, and the maximum depth from centimeters to millimeters, and write the calculation results into a structured data table.

[0061] Before entering the data entry process, the system performs a quality evaluation on the measurement results. The quality evaluation includes at least the quality of scale estimation, point cloud reconstruction, reference plane fitting, pit segmentation, and geometric result integrity. The quality of scale estimation can be characterized by the standard deviation, coefficient of variation, and number of effective viewpoints of the multi-view scale coefficients; the quality of reference plane fitting can be characterized by the proportion of points in the plane, the fitting residuals, and the normal stability; and the quality of pit segmentation can be characterized by the area of ​​connected regions, the continuity of depth distribution, and the proportion of outliers.

[0062] When the quality evaluation indicators do not meet the preset thresholds, the system may prompt the user to retake the photo, adjust the QR code position, increase the number of images, reselect the reference plane, or manually confirm the pit area. Measurement results that meet the quality evaluation standards or have been manually confirmed will enter the automatic reporting process. This quality evaluation result can be uploaded along with the inspection record for remote review and subsequent traceability.

[0063] Step S7: Automatically fill in, upload, and remotely view the detection results based on the QR code task information.

[0064] The system reads the URL or task identifier contained in the QR code, opens the corresponding online questionnaire, web form, inspection platform, or data interface, and automatically fills the corresponding fields with the pit geometry information obtained in step S6. Fields may include one or more of the following: pit location, pit size, average depth, maximum depth, volume, inspection time, inspection personnel, component number, pit photograph, and point cloud file.

[0065] In this implementation, the system uses the grid projection area as the pit size field and the maximum depth as the pit maximum depth field, and automatically performs unit conversion according to form requirements. The system can upload results via browser automation, platform interfaces, form interfaces, database writes, or mobile application calls. After uploading, remote administrators can view pit photos, geometric parameters, depth maps, point cloud files, and quality evaluation results, achieving digital storage and tracking management of concrete pit detection information.

[0066] Through the above steps, the present invention combines the scale correction function and information entry function of the visual scale marker, so that the multi-view image acquisition, three-dimensional reconstruction, real scale restoration, geometric measurement of the pit, automatic filling of detection results and remote viewing of concrete pits constitute a complete process.

[0067] System Implementation Examples: Figure 2 As shown, this embodiment provides a concrete pit measurement and reporting system based on dimensional markers, including: The image acquisition and labeling unit is used to deploy visual scale markers with known real side lengths near the concrete pit to be measured, and to guide the imaging equipment to acquire multi-view image sequences. The image quality control and identification unit is used to screen the quality of the acquired images, identify visual scale markers, and extract the corner pixel coordinates. The visual 3D reconstruction unit is used to generate scene point clouds, depth maps, and the correspondence between pixels and 3D points based on multi-view images; The scale correction unit is used to convert the reconstructed point cloud into a real-scale point cloud by robustly calculating the scale coefficient through multi-view statistical methods based on the three-dimensional coordinates of the corner points of the marker and the known true side length. The reference plane extraction and pit segmentation unit is used to preprocess the real-scale point cloud, fit the local reference plane, and segment and extract the point cloud of the pit region. The geometric information calculation unit is used to calculate the projected area, depth, volume, and contour parameters of the pit. The quality evaluation unit is used to evaluate the quality of scale estimation, point cloud reconstruction, reference plane fitting and pit segmentation, and output anomaly prompts. The automatic data entry and remote management unit is used to read the inspection task information carried by the visual scale marker, automatically fill in the measurement results and auxiliary data into the online form corresponding to the inspection task information, and synchronize it to the management platform for remote viewing and traceability.

[0068] The system can be deployed entirely on smartphones, or image acquisition can be deployed on smartphones while 3D reconstruction and computation are deployed on cloud servers, with data transmitted over a network. Individual units can also be deployed on mobile devices, edge computing devices, workstations, and cloud servers.

[0069] Besides QR codes, visual scale markers can also be replaced with other visual targets of known size, such as ArUco markers and AprilTag markers, all of which fall within the protection scope of this invention. Multi-view image acquisition can also display QR code corner detection results and image quality scores in real time during the acquisition process, allowing acquisition personnel to adjust their shooting posture on-site. The visual 3D reconstruction model can also combine semantic segmentation results to reconstruct only the concrete surface area, reducing interference from non-target objects in the measurement of pits.

[0070] The above are merely preferred embodiments of the present invention. The scope of protection of the present invention is not limited to the above embodiments. All technical solutions falling within the scope of the present invention's concept are within the scope of protection of the present invention. It should be noted that for those skilled in the art, any improvements and modifications made without departing from the principle of the present invention should also be considered within the scope of protection of the present invention.

Claims

1. A method for measuring and reporting concrete pits based on dimensional markers, characterized in that, Includes the following steps: S1. Visual scale markers are set up on the flat surface near the concrete pit to be tested. The visual scale markers have known true side lengths and carry detection task information. Multi-view image sequences containing the pit, the surrounding flat concrete surface and the visual scale markers are acquired from multiple perspectives by an imaging device. S2. Perform quality screening on the images in the multi-view image sequence, identify the visual scale markers in the screened images, and extract the corner pixel coordinates of the visual scale markers; S3. Input the selected multi-view images into the visual 3D reconstruction model to generate a scene point cloud containing pits, flat concrete surfaces and the visual scale markers, and establish the correspondence between image pixels and 3D points. S4. Obtain the corresponding three-dimensional point coordinates based on the corner pixel coordinates, calculate the side length of the visual scale marker in the reconstructed coordinate system, use the known real side length as a reference, combine the side length to obtain the scale coefficient through robust statistics, and use the scale coefficient to convert the reconstructed point cloud into a real scale point cloud. S5. Preprocess the real-scale point cloud, fit the local reference plane with the point cloud of the flat concrete surface around the pit, calculate the signed distance of each point in the point cloud to the local reference plane, and extract the continuous area below the local reference plane as the pit area. S6. Based on the point cloud of the pit region and the local reference plane, calculate the projected area, depth index, volume and contour parameters of the pit; S7. Read the detection task information carried by the visual scale marker, automatically fill in the projected area, depth index, volume and contour parameters of the pit and the acquired image into the online form corresponding to the detection task information, and synchronize it to the management platform for remote viewing and traceability.

2. The method for measuring and reporting concrete pits based on dimensional markers according to claim 1, characterized in that, The visual scale marker is a visual target, which can be any one of the following: a QR code, a square sign containing a QR code, a composite target with a pattern of known geometric dimensions, an ArUco logo, or an AprilTag logo; the detection task information includes the detection object number, component location, actual side length of the marker, form address, server address, detection task number, and timestamp.

3. The method for measuring and reporting concrete pits based on dimensional markers according to claim 1, characterized in that, In step S2, the quality screening includes sharpness evaluation, exposure evaluation, marker integrity evaluation, and image quantity evaluation. The sharpness evaluation is achieved by any one of Laplacian variance, edge strength, and depth reconstruction confidence. The marker integrity evaluation is achieved by decoding results, corner number, and corner sorting consistency. For images that fail to identify markers, their role in scale estimation is removed, but their right to participate in 3D reconstruction is retained.

4. The method for measuring and reporting concrete pits based on dimensional markers according to claim 1, characterized in that, In step S3, the visual 3D reconstruction model outputs a depth map, camera intrinsic parameters, camera pose, and world point map. The world point map is a 3D coordinate array that corresponds one-to-one with the image pixel coordinates, and is used to establish the correspondence between image pixels and 3D points.

5. The method for measuring and reporting concrete pits based on dimensional markers according to claim 1, characterized in that, In step S4, obtaining the scaling coefficients through robust statistics includes: For each available image, the reconstructed length of each side is calculated based on the three-dimensional coordinates of each corner point of the visual scale marker, and the scale candidate value of the image is obtained by combining the known true side length. The median, interquartile range, and median absolute deviation were used to remove outliers from the scale candidate values ​​of multiple images. After removing outlier viewpoints, the median of the remaining scale candidate values ​​was used as the final scale coefficient. It also outputs the standard deviation of the scale estimate, coefficient of variation, number of valid images, and number of outlier images. Converting the reconstructed point cloud into a true-scale point cloud using the scale coefficient includes: multiplying length-type indices by the scale coefficient, multiplying area-type indices by the square of the scale coefficient, and multiplying volume-type indices by the cube of the scale coefficient.

6. The method for measuring and reporting concrete pits based on dimensional markers according to claim 1, characterized in that, In step S5, the preprocessing includes statistical outlier filtering, radius outlier filtering, voxel downsampling, spatial range clipping, and voxel connected component analysis. The local reference plane is obtained by fitting any one of RANSAC, least squares fitting, robust plane fitting and region growth methods, and is used to characterize the original concrete surface before the pit occurs. The step of extracting continuous regions below the local reference plane as pit regions specifically involves: filtering candidate regions using depth thresholds, connectivity constraints, and region morphology constraints, and removing isolated noise points through radius filtering, retention of the largest connected component, and boundary smoothing.

7. The method for measuring and reporting concrete pits based on dimensional markers according to claim 1, characterized in that, In step S6, the projected area is calculated using any one of the following methods: regular grid method, AlphaShape, convex hull method, and boundary polygon method; the depth index includes maximum depth, average depth, and root mean square depth. The volume is calculated using the height map integral method: the projected area of ​​the pit is divided into a two-dimensional grid, and the mean, median and quantile of the pit depth in each grid are used as the representative depth of the grid. The volume is the sum of the products of the area of ​​each grid and the representative depth. The contour parameters include principal direction length, width, aspect ratio, perimeter, boundary contour, and depth heatmap.

8. The method for measuring and reporting concrete pits based on dimensional markers according to claim 1, characterized in that, Also includes: Before step S7, the quality of scale estimation, point cloud reconstruction, reference plane fitting, and pit segmentation is evaluated. The evaluation metrics include: The quality of scale estimation is characterized by the standard deviation, coefficient of variation, and number of effective views of the multi-view scale coefficients. Point cloud reconstruction quality is characterized by point cloud density, integrity, and reconstruction confidence. The fitting quality of the reference plane is characterized by the proportion of points in the plane, the fitting residuals, and the normal stability. The quality of pit segmentation is characterized by the area of ​​connected regions, the continuity of depth distribution, and the proportion of outliers. When the quality meets the standard, the system proceeds to the automatic data entry process in step S7; if the quality does not meet the standard, the system prompts for re-collection. Measurement results that have been manually confirmed proceed to the automatic data entry process in step S7.

9. A concrete pit measurement and reporting system based on dimensional markers, characterized in that, include: The image acquisition and labeling deployment unit is used to deploy visual scale markers near the concrete pit to be measured. The visual scale markers have known true side lengths and carry detection task information, and guide the imaging device to acquire multi-view image sequences. The image quality control and identification unit is used to screen the quality of the acquired images, identify the visual scale markers, and extract the corner pixel coordinates. The visual 3D reconstruction unit is used to generate scene point clouds, depth maps, and the correspondence between pixels and 3D points based on multi-view images; The scale correction unit is used to convert the reconstructed point cloud into a real-scale point cloud by calculating the scale coefficient through multi-view robust statistical calculation based on the three-dimensional coordinates of the corner points of the marker and the known true side length. The reference plane extraction and pit segmentation unit is used to preprocess the real-scale point cloud, fit the local reference plane, and segment and extract the point cloud of the pit region. The geometric information calculation unit is used to calculate the projected area, depth, volume, and contour parameters of the pit. The quality evaluation unit is used to evaluate the quality of scale estimation, point cloud reconstruction, reference plane fitting, and pit segmentation, and output anomaly prompts. The automatic reporting and remote management unit is used to read the detection task information carried by the visual scale marker, automatically fill in the measurement results and auxiliary data into the online form corresponding to the detection task information, and synchronize it to the management platform for remote viewing and traceability.

10. The concrete pit measurement and reporting system based on dimensional markers according to claim 9, characterized in that, The concrete pit measurement and filling system is deployed on the same computing terminal, or on a mobile terminal, edge computing device, workstation, and cloud server, and transmits data through a network; the computing terminal is any one of a mobile terminal, edge computing device, workstation, and cloud server; the imaging device is any one of a mobile phone, tablet computer, and drone equipped with a camera.

Images