Machine vision-based automatic identification and grading system for surface defects of shield segment

Abstract

The application discloses a kind of based on machine vision's shield segment surface defect automatic identification and grading system, it is related to shield segment detection technical field, when facing shield segment in solving the existing existing three-dimensional laser scanning technology, it cannot realize stable, reliable industrial quality control problem, the system includes mechanical arm three-dimensional laser scanning subsystem, segment automatic scanning subsystem and control and analysis subsystem;The mechanical arm three-dimensional laser scanning subsystem includes target bearing device, industrial robot and the laser scanning unit installed at the end of the industrial robot, wherein the laser scanning unit supports high frequency point cloud acquisition, adopts phase or triangular ranging principle, realizes millimeter level non-contact measurement, the industrial robot has six axes and above degrees of freedom, for according to scanning task planning motion path.The application has the advantages that the size deviation of the precise measurement of the precise identification of segment surface defect is realized.

Description

Technical Field

[0001] This invention relates to the field of tunnel boring machine segment inspection technology, and more specifically, to a machine vision-based automatic identification and grading system for surface defects in tunnel boring machine segments. Background Technology

[0002] With the booming development of ultra-large diameter shield tunnel projects, precast segments, as the core components of the main structure of tunnels, are directly related to the stability and safety of tunnels in terms of production precision. At present, the industry mainly relies on traditional tools such as steel rulers and vernier calipers for segment quality inspection. Manual measurement has drawbacks such as low efficiency, high operation difficulty, and accuracy being easily affected by subjective factors.

[0003] With the rise of 3D laser scanning technology, new technical means have been brought to industrial inspection. Its ranging principle covers pulse, phase and triangulation methods, which can meet different accuracy and distance requirements. Moreover, its system structure has developed into various forms such as fixed, vehicle-mounted and handheld, and is widely used in static precision measurement and dynamic environment modeling.

[0004] However, existing 3D laser scanning technology has significant shortcomings when it comes to inspecting large, irregularly shaped concrete components in tunnel lining segments. For example, fixed scanning systems cannot fully cover the complex curved surfaces of the segments due to limited field of view, while vehicle-mounted or mobile systems are easily affected by positioning errors and vibration interference when scanning small-scale components. Handheld devices are also susceptible to individual differences among operators. As a result, existing 3D laser scanning technology cannot achieve stable and reliable industrial-grade quality control when dealing with tunnel lining segments. In view of this, we propose an automatic identification and grading system for surface defects of tunnel lining segments based on machine vision. Summary of the Invention

[0005] The purpose of this invention is to provide an automatic identification and grading system for surface defects of tunnel boring machine segments based on machine vision, which aims to solve the problem that existing three-dimensional laser scanning technology cannot achieve stable and reliable industrial-grade quality control when dealing with tunnel boring machine segments.

[0006] To solve the above-mentioned technical problems, the present invention provides the following technical solution: an automatic identification and grading system for surface defects of tunnel segments based on machine vision, the system comprising a robotic arm three-dimensional laser scanning subsystem, an automatic segment scanning subsystem, and a control and analysis subsystem; The robotic arm 3D laser scanning subsystem includes a target support device, an industrial robotic arm, and a laser scanning unit installed at the end of the industrial robotic arm. The laser scanning unit supports high-frequency point cloud acquisition and uses phase-based or triangulation ranging principles to achieve millimeter-level non-contact measurement. The industrial robotic arm has six or more degrees of freedom and is used to plan motion paths according to the scanning task to perform high-density point cloud coverage on complex curved surfaces such as the inner and outer arc surfaces of the tube segment, bolt hole areas, and corner transition areas. The target support device is used to fix the tube segment to be measured to ensure the attitude stability and positioning consistency of the tube segment during the measurement process. The automatic tube segment scanning subsystem includes a laser tracking spatial measuring instrument, a gantry, an industrial robot, and a 3D laser scanner. The laser tracking spatial measuring instrument is designed to be suitable for the optimal scanning path of the tube segment. The gantry, by carrying the industrial robot and the 3D laser scanner, scans the formed tube segment and collects the tube segment point cloud data. The control and analysis subsystem integrates a motion planning module, a path control program, and a point cloud modeling algorithm. The control and analysis subsystem is used to coordinate the work of each subsystem, control the movement of the robotic arm and industrial robot, realize point cloud acquisition, data filtering, stitching, and 3D modeling, perform error comparison analysis against the standard design model, automatically identify cracks, edge damage, and dimensional deviation defects on the surface of the tube segment, and classify the quality according to the size of the defects.

[0007] Preferably, the laser scanning unit of the robotic arm's three-dimensional laser scanning subsystem includes a line laser scanner or an area laser scanner, with a scanning frequency of not less than 100Hz and a measurement accuracy of ±0.1mm.

[0008] Preferably, the laser tracking space measuring instrument of the automatic tube segment scanning subsystem has a measurement range of ≥10m, an angle measurement accuracy of ≤±2°, and a distance measurement accuracy of ≤±0.05mm.

[0009] Preferably, the point cloud modeling algorithm of the control and analysis subsystem includes a point cloud registration algorithm, a point cloud filtering algorithm, a surface reconstruction algorithm, and a feature extraction algorithm. The point cloud registration algorithm adopts the iterative nearest point algorithm with a registration accuracy of ≤ ±0.2mm. The point cloud filtering algorithm includes statistical filtering, radius filtering, or Gaussian filtering. The surface reconstruction algorithm adopts Poisson surface reconstruction or moving least squares method. The feature extraction algorithm is based on geometric features or deep learning features.

[0010] Preferably, the quality grading criteria of the control and analysis subsystem include classifying crack defects into minor cracks, moderate cracks, and severe cracks based on the length, width, and depth of the cracks; classifying corner damage defects into minor damage, moderate damage, and severe damage based on the area and depth of corner damage; and classifying dimensional deviation defects into acceptable, beyond-limit but repairable, and unacceptable based on the magnitude of the dimensional deviation.

[0011] Preferably, it also includes a defect marking and alarm module, which is used to visually mark the identified defects and trigger corresponding alarm signals according to the defect classification results. The alarm signals include audible and visual alarms, SMS alarms, or email alarms.

[0012] Preferably, it also includes a data management and traceability module for storing detection data, defect information, grading results and processing records, and supports data query, statistical analysis and historical traceability functions. The data management and traceability module uses a relational database or a non-relational database for data storage.

[0013] Preferably, the target support device includes an adjustable positioning clamp and a horizontal calibration mechanism, which can adapt to shield tunnel segments of different sizes and specifications, and ensure that the segment positioning error is ≤ ±0.5mm.

[0014] Compared with the prior art, the beneficial effects of the present invention are: This invention utilizes a robotic arm 3D laser scanning subsystem and an automatic segment scanning subsystem working in tandem. Through high-frequency point cloud acquisition, high-precision measurement, and rational path planning, it overcomes the limitations of fixed scanning field of view, interference from positioning and vibration in vehicle-mounted or mobile systems, and the influence of individual differences in handheld devices. This ensures complete and accurate scanning of the complex curved surfaces of the segments. Furthermore, the laser scanning unit, laser tracking spatial measuring instrument, and other components possess high-precision measurement parameters. The advanced point cloud modeling algorithm integrated into the control and analysis subsystem enables precise processing of the scanned data, reducing data errors and avoiding the accuracy deviations caused by subjective factors in manual measurement. This achieves accurate identification of surface defects and precise measurement of dimensional deviations in the segments.

[0015] This invention employs an automated scanning method, where robotic arms and industrial robots automatically scan tunnel segments along a preset path. Compared to traditional manual measurement using steel rulers and calipers, this significantly reduces manual operation time, accelerates the inspection process, and enables the rapid completion of inspections for a large number of tunnel segments.

[0016] The defect marking and alarm module in this invention can promptly and visually mark defects and issue graded alarms, facilitating quick location and handling of problems by staff. The data management and traceability module effectively stores and manages inspection data and defect information, supports data query, statistical analysis and historical traceability, and provides comprehensive data support and decision-making basis for segment quality control. Attached Figure Description

[0017] Figure 1 This is a schematic diagram of the system architecture in this invention. Detailed Implementation

[0018] The technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings of the embodiments of the present invention. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Example

[0019] like Figure 1 As shown, an automatic identification and grading system for surface defects of tunnel lining segments based on machine vision is provided. The system includes a robotic arm three-dimensional laser scanning subsystem, an automatic segment scanning subsystem, and a control and analysis subsystem. The robotic arm 3D laser scanning subsystem includes a target support device, an industrial robotic arm, and a laser scanning unit installed at the end of the industrial robotic arm. The laser scanning unit supports high-frequency point cloud acquisition and uses phase-based or triangulation ranging principles to achieve millimeter-level non-contact measurement. The industrial robotic arm has six or more degrees of freedom and is used to plan motion paths according to the scanning task to perform high-density point cloud coverage on complex curved surfaces such as the inner and outer arc surfaces of the tube segment, bolt hole areas, and corner transition areas. The target support device is used to fix the tube segment to be measured to ensure the attitude stability and positioning consistency of the tube segment during the measurement process. The automatic tube segment scanning subsystem includes a laser tracking spatial measuring instrument, a gantry, an industrial robot, and a 3D laser scanner. The laser tracking spatial measuring instrument is designed to be suitable for the optimal scanning path of the tube segment. The gantry, by carrying the industrial robot and the 3D laser scanner, scans the formed tube segment and collects the tube segment point cloud data. The control and analysis subsystem integrates a motion planning module, a path control program, and a point cloud modeling algorithm. The control and analysis subsystem is used to coordinate the work of each subsystem, control the movement of the robotic arm and industrial robot, realize point cloud acquisition, data filtering, stitching, and 3D modeling, perform error comparison analysis against the standard design model, automatically identify cracks, edge damage, and dimensional deviation defects on the surface of the tube segment, and classify the quality according to the size of the defects.

[0020] Specifically, the test segment is first fixed to the target support device, which uses its positioning and stabilization functions to ensure the segment's stable posture and consistent positioning throughout the measurement process. Next, the robotic arm's 3D laser scanning subsystem operates. The industrial robotic arm, with six or more degrees of freedom, moves along a pre-planned scanning path, driving the laser scanning unit at its end to scan complex curved surfaces such as the inner and outer arc surfaces, bolt hole areas, and corner transition areas of the segment. The laser scanning unit uses a high-frequency point cloud acquisition method based on phase-based or triangulation ranging principles to achieve millimeter-level non-contact measurement, acquiring point cloud data for relevant areas of the segment. Simultaneously, the automatic segment scanning subsystem begins operation. A laser tracking spatial measuring instrument designs the optimal scanning path for the segment. A gantry carrying an industrial robot and a 3D laser scanner scans the formed segment along this path, collecting overall point cloud data. Finally, the control and analysis subsystem coordinates the work of the first two subsystems, controlling the movement of the robotic arm and industrial robot, filtering and stitching the acquired point cloud data, and applying point cloud modeling algorithms. A 3D model of the tunnel segment is constructed, and then the error is compared and analyzed with the standard design model. Defects such as surface cracks, edge damage, and dimensional deviations of the tunnel segment are automatically identified, and quality classification is completed according to the size of the defects.

[0021] Furthermore, the laser scanning unit of the robotic arm's three-dimensional laser scanning subsystem includes a line laser scanner or an area laser scanner, with a scanning frequency of not less than 100Hz and a measurement accuracy of ±0.1mm.

[0022] Specifically, the laser scanning unit uses a line laser scanner or an area laser scanner, with a scanning frequency of not less than 100Hz and a measurement accuracy of ±0.1mm, to perform high-frequency and high-precision point cloud acquisition on the tube segment, thereby obtaining more accurate surface data of the tube segment and laying the foundation for more accurate identification and analysis of the surface condition of the tube segment in the future.

[0023] Furthermore, the laser tracking space measuring instrument of the automatic tube segment scanning subsystem has a measurement range of ≥10m, an angle measurement accuracy of ≤±2°, and a distance measurement accuracy of ≤±0.05mm.

[0024] Specifically, the laser tracking spatial measuring instrument, with a measurement range of ≥10m, an angle measurement accuracy of ≤±2°, and a distance measurement accuracy of ≤±0.05mm, designs the optimal scanning path. Subsequently, the gantry crane, equipped with an industrial robot and a 3D laser scanner, strictly follows this path to scan the tube segments. Thanks to the high-precision measurement capability of the laser tracking spatial measuring instrument, the point cloud data acquired is kept highly accurate even when measuring tube segments over a large area, providing reliable data support for subsequent analysis and processing.

[0025] Furthermore, the point cloud modeling algorithm of the control and analysis subsystem includes a point cloud registration algorithm, a point cloud filtering algorithm, a surface reconstruction algorithm, and a feature extraction algorithm. The point cloud registration algorithm adopts the iterative nearest point algorithm with a registration accuracy of ≤ ±0.2mm. The point cloud filtering algorithm includes statistical filtering, radius filtering, or Gaussian filtering. The surface reconstruction algorithm adopts Poisson surface reconstruction or moving least squares method. The feature extraction algorithm is based on geometric features or deep learning features.

[0026] Specifically, point cloud registration is performed first, using an iterative nearest-point algorithm to accurately register different point cloud data with a registration accuracy of ≤±0.2mm, achieving precise stitching of point cloud data. Next, point cloud filtering is performed, using statistical filtering, radius filtering, or Gaussian filtering algorithms to effectively remove noise points in the point cloud data and improve the quality of the point cloud data. Then, surface reconstruction is performed, using Poisson surface reconstruction or moving least squares method to transform the processed point cloud data into a three-dimensional surface model of the tunnel segment. Finally, surface feature information of the tunnel segment is extracted from the reconstructed surface model through geometric feature-based or deep learning feature extraction algorithms, providing an important basis for subsequent accurate identification of surface defects of the tunnel segment.

[0027] Furthermore, the quality grading criteria of the control and analysis subsystem include classifying crack defects into minor cracks, moderate cracks, and severe cracks based on the length, width, and depth of the cracks; classifying corner damage defects into minor damage, moderate damage, and severe damage based on the area and depth of the corner damage; and classifying dimensional deviation defects into acceptable, beyond-limit but repairable, and unacceptable based on the magnitude of the dimensional deviation.

[0028] Specifically, crack defects are classified into three levels—minor cracks, moderate cracks, and severe cracks—based on their length, width, and depth. Corner damage defects are classified into three levels—minor damage, moderate damage, and severe damage—based on the area and depth of the damage. Dimensional deviation defects are classified into three levels—acceptable, beyond limits but repairable, and unacceptable—based on the size of the dimensional deviation. Through such clear and detailed classification standards, the condition of tunnel segment defects can be systematically and accurately assessed.

[0029] Furthermore, it also includes a defect marking and alarm module, which is used to visually mark the identified defects and trigger corresponding alarm signals based on the defect classification results. The alarm signals include audible and visual alarms, SMS alarms, or email alarms.

[0030] Specifically, the defect marking and alarm module visually marks the identified defects on the segment model or actual segment image, enabling staff to intuitively and clearly view the location and condition of the defects. At the same time, based on the defect classification results, it triggers corresponding alarm signals, including audible and visual alarms, SMS alarms, or email alarms, to promptly notify relevant personnel of the defect information so that they can take swift measures to address the defects.

[0031] Furthermore, it also includes a data management and traceability module for storing detection data, defect information, grading results and processing records, and supports data query, statistical analysis and historical traceability functions. The data management and traceability module uses a relational database or a non-relational database for data storage.

[0032] Specifically, the data management and traceability module stores inspection data, defect information, grading results, and processing records using either a relational or non-relational database. This data storage method allows users to perform data queries, statistical analysis, and historical traceability operations, facilitating in-depth review and analysis of the segment quality inspection results. This helps in summarizing experience, optimizing inspection processes, and improving segment quality.

[0033] Furthermore, the target support device includes an adjustable positioning clamp and a horizontal calibration mechanism, which can adapt to shield tunnel segments of different sizes and specifications, and ensure that the segment positioning error is ≤ ±0.5mm.

[0034] Specifically, adjustable positioning fixtures and a horizontal calibration mechanism are used to precisely position and calibrate tunnel segments of different sizes, strictly controlling the segment positioning error within ≤±0.5mm. This paves the way for subsequent 3D laser-guided robotic arms. The accurate measurement and scanning operations of the scanning subsystem and the automatic segment scanning subsystem provide a stable and reliable foundation, ensuring the accuracy and effectiveness of the entire inspection process.

[0035] The embodiments disclosed in this invention are preferred embodiments, but are not limited thereto. Those skilled in the art can easily understand the spirit of this invention based on the above embodiments and make different extensions and variations, but as long as they do not depart from the spirit of this invention, they are all within the protection scope of this invention.

Claims

1. A machine vision-based automatic identification and grading system for surface defects in tunnel boring machine segments, characterized in that, The system includes a robotic arm 3D laser scanning subsystem, an automatic segment scanning subsystem, and a control and analysis subsystem; The robotic arm 3D laser scanning subsystem includes a target support device, an industrial robotic arm, and a laser scanning unit installed at the end of the industrial robotic arm. The laser scanning unit supports high-frequency point cloud acquisition and uses phase-based or triangulation ranging principles to achieve millimeter-level non-contact measurement. The industrial robotic arm has six or more degrees of freedom and is used to plan motion paths according to the scanning task to perform high-density point cloud coverage on complex curved surfaces such as the inner and outer arc surfaces of the tube segment, bolt hole areas, and corner transition areas. The target support device is used to fix the tube segment to be measured to ensure the attitude stability and positioning consistency of the tube segment during the measurement process. The automatic tube segment scanning subsystem includes a laser tracking spatial measuring instrument, a gantry, an industrial robot, and a 3D laser scanner. The laser tracking spatial measuring instrument is designed to be suitable for the optimal scanning path of the tube segment. The gantry, by carrying the industrial robot and the 3D laser scanner, scans the formed tube segment and collects the tube segment point cloud data. The control and analysis subsystem integrates a motion planning module, a path control program, and a point cloud modeling algorithm. The control and analysis subsystem is used to coordinate the work of each subsystem, control the movement of the robotic arm and industrial robot, realize point cloud acquisition, data filtering, stitching, and 3D modeling, perform error comparison analysis against the standard design model, automatically identify cracks, edge damage, and dimensional deviation defects on the surface of the tube segment, and classify the quality according to the size of the defects.

2. The automatic identification and grading system for surface defects of tunnel lining segments based on machine vision according to claim 1, characterized in that, The laser scanning unit of the robotic arm's three-dimensional laser scanning subsystem includes a line laser scanner or an area laser scanner, with a scanning frequency of not less than 100Hz and a measurement accuracy of ±0.1mm.

3. The automatic identification and grading system for surface defects of tunnel lining segments based on machine vision according to claim 1, characterized in that, The laser tracking space measuring instrument of the automatic tube segment scanning subsystem has a measurement range of ≥10m, an angle measurement accuracy of ≤±2°, and a distance measurement accuracy of ≤±0.05mm.

4. The automatic identification and grading system for surface defects of tunnel lining segments based on machine vision according to claim 1, characterized in that, The point cloud modeling algorithm of the control and analysis subsystem includes a point cloud registration algorithm, a point cloud filtering algorithm, a surface reconstruction algorithm, and a feature extraction algorithm. The point cloud registration algorithm adopts the iterative nearest point algorithm with a registration accuracy of ≤ ±0.2mm. The point cloud filtering algorithm includes statistical filtering, radius filtering, or Gaussian filtering. The surface reconstruction algorithm adopts Poisson surface reconstruction or moving least squares method. The feature extraction algorithm is based on geometric features or deep learning features.

5. The automatic identification and grading system for surface defects of tunnel lining segments based on machine vision according to claim 1, characterized in that, The quality grading standards of the control and analysis subsystem include classifying crack defects into minor cracks, moderate cracks, and severe cracks based on the length, width, and depth of the cracks; classifying corner damage defects into minor damage, moderate damage, and severe damage based on the area and depth of corner damage; and classifying dimensional deviation defects into acceptable, beyond-limit but repairable, and unacceptable based on the magnitude of the dimensional deviation.

6. The automatic identification and grading system for surface defects of tunnel lining segments based on machine vision according to claim 1, characterized in that, It also includes a defect marking and alarm module, which is used to visually mark the identified defects and trigger corresponding alarm signals based on the defect classification results. The alarm signals include audible and visual alarms, SMS alarms, or email alarms.

7. The automatic identification and grading system for surface defects of tunnel lining segments based on machine vision according to claim 1, characterized in that, It also includes a data management and traceability module, which stores detection data, defect information, grading results and processing records, and supports data query, statistical analysis and historical traceability functions. The data management and traceability module uses a relational database or a non-relational database for data storage.

8. The automatic identification and grading system for surface defects of tunnel lining segments based on machine vision according to claim 1, characterized in that, The target support device includes an adjustable positioning clamp and a horizontal calibration mechanism, which can adapt to shield tunnel segments of different sizes and specifications, and ensure that the segment positioning error is ≤ ±0.5mm.

Images