Tunnel crack image acquisition and detection device and detection method

Abstract

The invention provides a tunnel crack image acquisition and detection device and method, and the device comprises a rail car which is provided with a distance detection mechanism; the image acquisition device comprises an image acquisition module and a steering adjustment module, the image acquisition module is mounted at the top of the rail car through the steering adjustment module, and the steering adjustment module is used for driving the image acquisition module to steer so as to adjust the orientation of the image acquisition module; the crack identification device is arranged at the front end of the image acquisition module and is used for identifying the crack and determining the position of the crack; and the control module is electrically connected with the steering adjustment module, the crack recognition device and the distance detection mechanism, and is used for controlling the steering adjustment module to drive the image acquisition module to steer after the crack is recognized, so that the image acquisition module is aligned with the crack position. The device provided by the invention is simple in structure and low in maintenance cost, does not need to configure a plurality of industrial cameras and lenses, and saves the system cost.

Description

Technical Field

[0001] The present invention generally relates to the technical field of tunnel crack detection, and in particular to a tunnel crack image acquisition and detection device. Background Art

[0002] Cracks are a common early manifestation of tunnel damage, often appearing in tunnel linings. Failure to promptly detect and repair cracks in the lining can compromise the lining's load-bearing capacity and safety factor, shortening the tunnel's service life and potentially leading to surface delamination and other issues, compromising tunnel safety. Consequently, numerous researchers, both domestically and internationally, have conducted extensive research in the field of tunnel lining crack detection, aiming to design a simple and efficient system to meet the growing demand for tunnel inspection.

[0003] Technically, existing inspection systems typically collect images of the entire lining surface of a tunnel section, with data volumes reaching up to 40GB per kilometer. However, cracks often only represent a small fraction of the lining surface, resulting in a significant amount of useless information in the collected data, making image processing and target recognition algorithms more challenging. In terms of cost, systems based on large rail vehicles are both expensive to develop and maintain. While systems based on small inspection vehicles are significantly less expensive than large rail vehicles, capturing images of the entire tunnel lining cross-section typically requires more than six industrial cameras and lenses, increasing system costs. Summary of the Invention

[0004] In view of the above-mentioned defects or deficiencies in the prior art, it is desired to provide a tunnel crack image acquisition and detection device and detection method to solve the above-mentioned problems.

[0005] A first aspect of the present invention provides a tunnel crack image acquisition and detection device, comprising: A railcar, wherein the railcar is provided with a distance detection mechanism for detecting the travel distance of the railcar; An image acquisition device, the image acquisition device comprising an image acquisition module and a steering adjustment module, the image acquisition module being mounted on top of the railcar via the steering adjustment module, the steering adjustment module being configured to drive the image acquisition module to steer to adjust the orientation of the image acquisition module; A crack identification device, which is provided at the front end of the image acquisition module and is used to identify cracks and determine crack orientations; A control module is electrically connected to the steering adjustment module, the crack identification device and the distance detection mechanism. The control module is used to control the steering adjustment module to drive the image acquisition module to steer after identifying a crack, so that the image acquisition module is aligned with the crack position.

[0006] According to the technical solution provided by the present invention, the steering adjustment module includes: a graduation support assembly connected to the image acquisition module, the graduation support assembly comprising a first rotation axis parallel to the top surface of the rail vehicle; the graduation support assembly is used to drive the image acquisition module to rotate around the first rotation axis; A bottom rotating assembly connects the rail car and the indexing support assembly, the bottom rotating assembly includes a second rotating axis, and the second rotating axis is perpendicular to the top surface of the rail car; the bottom rotating assembly is used to drive the indexing support assembly to rotate around the second rotating axis.

[0007] According to the technical solution provided by the present invention, the indexing support assembly further includes: a pair of support plates, the pair of support plates being arranged along the extension direction of the first rotating axis and mounted on the indexing support assembly, the support plates being fan-shaped and having an arc-shaped first notch, the inner edge of the first notch being a V-shaped guide rail; a pair of connecting members, the pair of connecting members being arranged between the pair of support plates, one end of the connecting member being fixed to the first rotating shaft, and the other end of the connecting member being fixed to the image acquisition module; a guide plate being provided on the connecting member, a plurality of V-shaped wheels being rotatably mounted on the guide plate, and the V-shaped wheels being mounted on the V-shaped guide rails on corresponding sides; A first driving assembly is connected to the first rotating shaft in a transmission manner. The first driving assembly is installed on the support plate and is electrically connected to the control module, and is used to drive the first rotating shaft to rotate.

[0008] According to the technical solution provided by the present invention, the bottom rotating assembly further includes: an upper rotating disk, the upper rotating disk being connected to the indexing support assembly via an upper bracket; a lower rotating plate, the lower rotating plate being mounted on the top of the rail vehicle via a lower bracket, the lower rotating plate and the upper rotating plate being rotatable relative to each other about the second rotating axis so as to switch the upper rotating plate between a first station and a second station, the first station and the second station being angularly spaced 180° apart; a locking assembly, the locking assembly being mounted on the upper rotating disk and the lower rotating disk, respectively, and being used to lock the upper rotating disk at the first station or the second station; A second driving assembly is connected to the second rotating shaft in a transmission manner. The second driving assembly is installed on the lower rotating disk and is electrically connected to the control module, and is used to drive the second rotating shaft to rotate.

[0009] According to the technical solution provided by the present invention, the image acquisition module includes: An arc-shaped housing connected to the indexing support assembly, a light source and an area array camera being provided on the top of the arc-shaped housing, the area array camera being used to capture images of the cracks; A plurality of laser infrared emitters are evenly distributed around the arc-shaped shell, and the laser infrared emitters are connected to the arc-shaped shell through a damping shaft.

[0010] According to the technical solution provided by the present invention, the crack identification device includes: an identification camera, the identification camera being rotatably mounted on the support plate near the front end of the railcar, with a rotation axis coaxial with the first rotation axis; the identification camera being used to identify cracks, the identification camera being electrically connected to the control module; The third driving component, the second driving component and the recognition camera are installed on the same support plate, the output end of the third driving component is connected to the recognition camera in a transmission manner, and the third driving component is used to drive the recognition camera to rotate.

[0011] A second aspect of the present invention provides a method for collecting and detecting tunnel crack images, which is applied to the above-mentioned tunnel crack image collection and detection device. The method comprises: The railcar moves along the tunnel a set distance each time to an identification point; The crack identification device identifies the crack at the identification point, and sends the crack orientation to the control module after identifying the crack; The control module controls the steering adjustment module to adjust the image acquisition module toward the crack according to the crack orientation, and obtains the travel distance of the rail vehicle; The image acquisition module acquires crack images and calculates actual sizes of the cracks according to the crack images, so that the control module generates complete crack information according to the actual sizes of the cracks, the travel distance and the crack orientation.

[0012] According to the technical solution provided by the present invention, the image acquisition module acquires crack images and calculates the actual size of the cracks based on the crack images, including: Acquire the crack image, perform crack target detection on the crack image, and locate a rectangular bounding box area of the crack; An improved Panoptic FPN network is used to perform fine segmentation of crack pixels to obtain a segmented image; The segmented image is thinned to generate a crack skeleton, and the crack length and width are calculated based on the crack skeleton.

[0013] According to the technical solution provided by the present invention, the improved Panoptic FPN network includes: The encoder adopts the ResNeSt multi-level feature extractor; the decoder upsamples and concatenates multi-scale features through a feature pyramid network, and uses the decoder head to predict and output the segmentation results of crack pixels.

[0014] According to the technical solution provided by the present invention, the thinning process of the segmented image to generate a crack skeleton, and the calculation of the crack length and width based on the crack skeleton includes: The crack skeleton is obtained using the Zhang-Suen thinning algorithm; Calculating the sum of the distances between adjacent pixels on the crack skeleton to obtain the pixel-level length of the crack, and using the local perpendicular line method to obtain the pixel-level width of the crack; Combined with Zhang Zhengyou's calibration method, the actual length and width of the crack are calculated.

[0015] Compared with the prior art, the beneficial effects of the present invention are as follows: through the collaborative design of the integrated steering adjustment module and the single image acquisition module, there is no need to arrange multiple industrial cameras and lenses on the rail car, which significantly reduces the hardware complexity and system cost; by setting up a crack recognition device, cracks in the tunnel can be identified, and the control module drives the steering adjustment module in real time to adjust the orientation of the image acquisition module based on the crack orientation feedback from the crack recognition device, so that it is accurately aligned with the crack position, thereby improving the efficiency of crack image acquisition and further improving the efficiency of crack size calculation; by adopting standardized electrical connections for each module component, the overall structure of the device is compact, and standardized electrical connections are adopted between modules, which reduces mechanical redundant components, reduces the failure rate, and facilitates daily maintenance and upgrades. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] Other features, objects and advantages of the present invention will become more apparent upon reading the detailed description of non-limiting embodiments made with reference to the following drawings: Figure 1 A schematic structural diagram of the tunnel crack image acquisition and detection device provided by the present invention; Figure 2 for Figure 1 The front-end schematic diagram of the tunnel crack image acquisition and detection device shown; Figure 3 It is a structural diagram of the indexing support assembly; Figure 4 for Figure 3 a side view of the indexing support assembly shown; Figure 5 It is a structural diagram of the bottom rotating assembly; Figure 6 It is a structural schematic diagram of the upper rotating disk and the lower rotating disk; Figure 7 is a cross-sectional view of the upper rotating disk and the lower rotating disk; Figure 8 It is a structural diagram of the image acquisition module; Figure 9 This is the crack target detection result diagram; Figure 10 To improve the FPN network structure diagram; Figure 11 Schematic diagram of the image before and after processing by Zhang_Suen algorithm; Figure 12 Schematic diagram of the 3×3 area of a pixel on the crack skeleton; Figure 13 Schematic diagram of the possible local vertical line direction of a point on the column skeleton.

[0017] Reference numerals: 100, railcar; 110, distance detection mechanism; 200, image acquisition module; 210, arc-shaped housing; 220, light source; 230, area array camera; 240, laser infrared emitter; 300, steering adjustment module; 310, indexing support assembly; 311, first rotation axis; 312, support plate; 313, first notch; 314, second notch; 315, connector; 316, guide plate; 317, V-shaped wheel; 318, bottom plate; 319, connecting plate; 3 20. Bottom rotating assembly; 321. Second rotating shaft; 322. Upper rotating disk; 323. Upper bracket; 324. Lower rotating disk; 325. Lower bracket; 331. Spring positioning bead; 332. Positioning hole; 340. Proximity switch; 350. Magnet; 400. Crack recognition device; 410. Recognition camera; 411. Steering connecting rod; 420. Third driving assembly; 510. First servo motor; 520. First reducer; 530. Second servo motor; 540. Second reducer. DETAILED DESCRIPTION

[0018] The present invention will be further described in detail below with reference to the accompanying drawings and examples. It should be understood that the specific embodiments described herein are intended only to illustrate the relevant invention and are not intended to limit the invention. It should also be noted that, for ease of description, only portions relevant to the invention are shown in the accompanying drawings.

[0019] It should be noted that, in the absence of conflict, the embodiments and features of the embodiments of the present invention can be combined with each other. The present invention will be described in detail below with reference to the accompanying drawings and in combination with the embodiments.

[0020] Example 1 Please refer to Figures 1-8 This embodiment provides a tunnel crack image acquisition and detection device, including: A railcar 100 is provided with a distance detection mechanism 110 for detecting a travel distance of the railcar 100; An image acquisition device, comprising an image acquisition module 200 and a steering adjustment module 300. The image acquisition module 200 is mounted on top of the rail vehicle 100 via the steering adjustment module 300. The steering adjustment module 300 is used to drive the image acquisition module 200 to steer to adjust the orientation of the image acquisition module 200. A crack identification device 400, which is provided at the front end of the image acquisition module 200 and is used to identify cracks and determine crack orientations; A control module is electrically connected to the steering adjustment module 300, the crack identification device 400 and the distance detection mechanism 110. The control module is used to control the steering adjustment module 300 to drive the image acquisition module to steer after identifying a crack, so that the image acquisition module 200 is aligned with the crack position.

[0021] Specifically, the tunnel crack image acquisition and detection device provided in this embodiment is composed of a hardware system and a software system. The hardware system includes a rail vehicle 100, an image acquisition device, a crack identification device 400, and a control module.

[0022] The track car 100 is used to move the entire device within the tunnel. The track car 100 can be a small, hand-pushed track car or an automatically advancing track car. A distance detection mechanism 110 is provided on the track car 100. This distance detection mechanism 110 can detect the distance traveled by the track car 100, facilitating the subsequent determination of the crack's location within the tunnel based on the distance traveled by the track car. In this embodiment, the distance detection mechanism 110 is an encoder mounted on the wheels of the track car 100. The encoder detects the rotation angle of the wheels of the track car 100 and calculates the distance traveled based on the wheel circumference. The image acquisition device, crack identification device 400, and control module are all mounted on the top of the track car 100. In addition, an industrial computer and a power supply are also installed on the track car 100. The industrial computer facilitates human-computer interaction; the specific structures of the industrial computer and power supply are not shown in the accompanying drawings. The power supply is connected to the image acquisition device, crack identification device 400, and control module to power each module.

[0023] The image acquisition device consists of an image acquisition module 200 and a steering adjustment module 300. The image acquisition module 200 has a high-precision image acquisition function and is used to acquire high-definition crack images so that the size of the cracks can be calculated based on the crack images. The steering adjustment module 300 is used to connect the image acquisition module 200 to the railcar 100, and the steering adjustment module 300 can adjust the image acquisition module 200 at multiple angles so that crack images at any angle in the tunnel can be acquired using a single image acquisition module 200. The crack identification device 400 is installed near the front end of the railcar 100. The accuracy of the crack identification device 400 is lower than that of the image acquisition module 200. It scans the inner wall of the tunnel to identify whether there are cracks. Every time the rail vehicle 100 moves forward a certain distance and stops, the crack recognition device 400 first scans the corresponding position in the tunnel. When the crack recognition device 400 recognizes a crack, it sends the crack orientation to the control module. The control module controls the steering adjustment module 300 according to the crack orientation. The steering adjustment module 300 operates to adjust the direction of the image acquisition module 200 and align it with the crack. After the image acquisition module 200 is aligned with the crack, it captures the crack image. The crack image is then processed according to the software system to calculate the specific size of the crack.

[0024] Furthermore, the steering adjustment module 300 includes: An indexing support assembly 310 is connected to the image acquisition module 200 and includes a first rotation axis 311 parallel to the top surface of the rail vehicle 100. The indexing support assembly 310 is configured to drive the image acquisition module 200 to rotate about the first rotation axis 311. The bottom rotating assembly 320 connects the rail vehicle 100 and the indexing support assembly 310, and the bottom rotating assembly 320 includes a second rotating shaft 321, which is perpendicular to the top surface of the rail vehicle 100; the bottom rotating assembly 320 is used to drive the indexing support assembly 310 to rotate around the second rotating shaft 321.

[0025] Specifically, the steering adjustment module 300 consists of a graduated support assembly 310 and a bottom rotating assembly 320. The graduated support assembly 310 supports the image acquisition module 200 and adjusts the orientation of the image acquisition module 200 by driving the image acquisition module 200 to rotate about a first rotating axis 311. The bottom rotating assembly 320 supports the graduated support assembly 310 and drives the graduated support assembly 310 to rotate about a second rotating axis 321, thereby increasing the adjustment range of the image acquisition module 200 and achieving multi-dimensional adjustment of the orientation of the image acquisition module 200.

[0026] Furthermore, the indexing support assembly 310 further includes: A pair of support plates 312, the pair of support plates 312 are arranged along the extension direction of the first rotating shaft 311 and mounted on the indexing support assembly 310, the support plates 312 are fan-shaped and have arc-shaped first notches 313, and the inner edges of the first notches 313 are V-shaped guide rails; A pair of connecting members 315 are provided between the pair of support plates 312, one end of each connecting member 315 is fixed to the first rotating shaft 311, and the other end of each connecting member 315 is fixed to the image acquisition module 200; a guide plate 316 is provided on each connecting member 315, and a plurality of V-shaped wheels 317 are rotatably mounted on each guide plate 316, and each V-shaped wheel 317 is mounted on the V-shaped guide rail on the corresponding side; A first driving assembly is connected to the first rotating shaft 311 in a transmission manner. The first driving assembly is installed on the supporting plate 312 and is electrically connected to the control module, and is used to drive the first rotating shaft 311 to rotate.

[0027] Specifically, the bottoms of the pair of support plates 312 are connected by a base plate 318, which is connected to the bottom rotating assembly 320. One side of the pair of support plates 312 is connected and fixed by a connecting plate 319, and the other side is arc-shaped. The pair of support plates 312 are penetrated by a first rotating shaft 311, which is driven by a first driving assembly. The first driving assembly is mounted on one of the support plates 312 and includes a first servo motor 510 and a first reducer 520. The first rotating shaft 311 is in transmission connection with the first reducer 520. A pair of connecting members 315 are provided on the inner sides of the pair of support plates 312. The pair of connecting members 315 are respectively arranged close to the pair of support plates 312 and extend in the radial direction of the support plates 312. One end of the pair of connecting members 315 is fixedly connected to the first rotating shaft 311, and the other end is connected to the image acquisition module 200. The connecting member 315 is limited and guided by a pair of support plates 312. When the first driving assembly drives the first rotating shaft 311 to rotate, the first rotating shaft 311 drives the connecting member 315 to rotate, thereby driving the image acquisition module 200 to adjust its direction around the first rotating shaft 311.

[0028] To enhance the limiting and guiding effect of the support plate 312 on the connector 315, a first arcuate notch 313 is provided on one side of the arcuate support plate 312. The first notch 313 penetrates the support plate 312, and the inner edge of the first notch 313 projects into the groove to form a V-shaped guide rail. A guide plate 316 is provided on the connector 315, which fits in the first notch 313. A plurality of V-shaped wheels 317 are provided on the side of the guide plate 316 away from the connector 315. The V-shaped wheels 317 are engaged with the V-shaped guide rail. During the rotation of the connector 315, the V-shaped wheels 317 roll on the V-shaped guide rail, thereby improving the stability of the connection between the connector 315 and the support plate 312 and enhancing the limiting and guiding effect of the support plate 312 on the connector 315.

[0029] In this embodiment, the first notch 313 limits the rotation angle of the connecting member 315 to 110°, allowing the indexing support assembly 310 to adjust the image acquisition module 200 to capture only half of the tunnel lining surface image. This design is intended to reduce the overall mass of the steering adjustment module 300. To achieve the acquisition of images of the entire tunnel lining surface, the indexing support assembly 310 can cooperate with the bottom rotation assembly 320.

[0030] Furthermore, the bottom rotating assembly 320 further includes: An upper rotating disk 322 , the upper rotating disk 322 being connected to the indexing support assembly 310 via an upper bracket 323 ; A lower rotating disk 324 is mounted on the top of the rail vehicle 100 via a lower bracket 325. The lower rotating disk 324 and the upper rotating disk 322 are relatively rotatable around the second rotating axis 321, so that the upper rotating disk 322 can switch between a first station and a second station. The first station and the second station have an angular difference of 180 degrees. A locking assembly, the locking assembly being mounted on the upper rotating disk 322 and the lower rotating disk 324 respectively, and the locking assembly being used to lock the upper rotating disk 322 at the first station or the second station; The second driving assembly is in transmission connection with the second rotating shaft 321 . The second driving assembly is installed on the lower rotating disk 324 and is electrically connected to the control module, and is used to drive the second rotating shaft 321 to rotate.

[0031] Specifically, the bottom rotating assembly 320 is mainly composed of an upper rotating disk 322, a lower rotating disk 324, and a second drive assembly. The upper rotating disk 322 and the lower rotating disk 324 can rotate relative to each other. Four mounting holes are set at the top of the upper rotating disk 322 for connecting to the bottom of the upper bracket 323; four mounting holes are set at the bottom of the lower rotating disk 324 for connecting to the top of the lower bracket 325. The top of the upper bracket 323 is fixedly connected to the base plate 318, and the bottom of the upper bracket 323 is fixedly connected to the top of the rail car 100. The second rotating shaft 321 passes through the lower rotating disk 324 and is fixedly connected to the upper rotating disk 322 at one end. The second rotating shaft 321 is driven by the second drive assembly, which is mounted on the lower rotating disk 322. The second drive assembly includes a second servo motor 530 and a second reducer 540. The end of the second rotating shaft 321 away from the upper rotating disk 322 is transmission-connected to the second reducer 540. When the second driving assembly drives the second rotating shaft 321 to rotate, the second rotating shaft 321 drives the upper rotating disk 322 to rotate, thereby driving the entire indexing support assembly 310 to rotate, thereby driving the image acquisition module 200 to adjust its direction around the second rotating shaft 321 .

[0032] Since cracks exist on the tunnel lining surface, when the rail car 100 advances in the tunnel, the dividing support assembly 310 can already capture the tunnel lining surface image on one side of the tunnel. Therefore, the dividing support assembly 310 is rotated 180° to capture the complete tunnel lining surface image. Therefore, the upper rotating disk 322 only needs to rotate 180° relative to the lower rotating disk 324. Therefore, the upper rotating disk 322 is set with a first workstation and a second workstation, and the rotation of the upper rotating disk 322 only switches between the first workstation and the second workstation.

[0033] The indexing support assembly 310 can rotate 110° around the tunnel's central axis. Combined with the bottom rotation assembly 320, it drives a 180° rotation of the indexing support assembly 310, giving the steering adjustment module 300 a 220° adjustment range, ensuring full coverage of the tunnel's inspection area. Compared to a single-degree-of-freedom rotation of 220°, this design reduces the mass of the steering adjustment module 300.

[0034] The locking assembly includes two spring-loaded positioning beads 331 and two positioning holes 332. The two spring-loaded positioning beads 331 are installed on the side of the upper rotating disk 322 near the lower rotating disk 324, with the two spring-loaded positioning beads 331 spaced 180 degrees apart. The two positioning holes 332 are located on the side of the lower rotating disk 324 near the upper rotating disk 322, corresponding to the two spring-loaded positioning beads 331. The cooperation between the spring-loaded positioning beads 331 and the positioning holes 332 allows the upper rotating disk 322 to be locked in both the first and second positions.

[0035] Specifically, a proximity switch 340 is provided on one side of the lower rotating disk 324, and a magnet 350 for triggering the proximity switch 340 is provided on the upper rotating disk 322. The state of the proximity switch 340 is used to determine whether the upper rotating disk 322 is in the first position or the second position.

[0036] Furthermore, the image acquisition module 200 includes: An arc-shaped housing 210, the arc-shaped housing 210 being connected to the indexing support assembly 310, a light source 220 and an area array camera 230 being provided on the top of the arc-shaped housing 210, the area array camera 230 being used to capture images of the cracks; A plurality of laser infrared emitters 240 are evenly distributed around the arc-shaped housing 210 , and the laser infrared emitters 240 are connected to the arc-shaped housing 210 via a damping shaft.

[0037] Specifically, the curved housing 210 is secured to a pair of connectors 315 on either side via bolts. A light source 220 and an area array camera 230 are mounted on top of the curved housing 210 to capture crack images. The area array camera 230 features high-precision image acquisition. The light source 220 illuminates the crack location to increase brightness, thereby improving the quality of the crack image. Multiple laser infrared emitters 240 are also positioned around the curved housing 210. In this embodiment, four laser infrared emitters 240 are provided, mounted on each of the four sides of the curved housing 210. The laser infrared emitters 240 are connected to the curved housing 210 via a damping shaft. The laser infrared emitters 240 emit a line of laser light. By adjusting the damping shaft to the appropriate position, the area formed by the four laser beams can be aligned with the field of view of the area array camera 230, facilitating real-time viewing of the camera's field of view for the operator. A battery is also located within the curved housing 210 to power the laser infrared emitters 240.

[0038] Furthermore, the crack identification device 400 includes: an identification camera 410 rotatably mounted on the support plate 312 near the front end of the railcar 100 , with its rotation axis coaxial with the first rotation axis 311 ; the identification camera 410 is used to identify cracks, and the identification camera 410 is electrically connected to the control module; The third driving component 420, the second driving component and the recognition camera 410 are installed on the same support plate 312, the output end of the third driving component 420 is transmission-connected to the recognition camera 410, and the third driving component 420 is used to drive the recognition camera 410 to rotate.

[0039] Specifically, the recognition camera 410 is mounted on the support plate 312 via a steering rod 411. The steering rod 411 suspends the recognition camera 410, and one end of the steering rod 411 is rotatably connected to the support plate 312, thereby enabling the recognition camera 410 to rotate relative to the support plate 312. The recognition camera 410 has lower accuracy than the area array camera 230. The recognition camera 410 is used to detect cracks in the tunnel lining, without the need to obtain high-definition images of cracks. A third drive assembly 420 is used to drive the recognition camera 410 in rotation. Like the first drive assembly, the third drive assembly 420 consists of a servo motor and a reducer, which is in transmission connection with the steering rod 411. Driven by the third drive assembly 420, the recognition camera 410 can rotate 360° about the first rotation axis 311.

[0040] In order to enhance the support and guidance of the recognition camera 410 when turning, a second slot 314 is also provided on the support plate 312. The second slot 314 is also arc-shaped. The structure of the second slot 314 is similar to that of the first slot 313. The recognition camera 410 and the second slot 314 are also connected and matched through the V-shaped wheel 317 and the V-shaped guide rail.

[0041] Every time the rail car 100 moves forward a certain distance and stops, the servo motor drives the steering connecting rod 411 to rotate one circle, so that the recognition camera 410 can identify whether there is a crack on the circumferential lining at that position. After the crack is identified, the position of the crack in the circumferential direction of the lining can be determined based on the rotation angle of the servo motor. Then, the image acquisition module 200 is directed to the crack position through the indexing support assembly 310 and the bottom rotating assembly 320 to collect the crack image, thereby realizing automatic recognition and image collection of the crack.

[0042] Example 2 This embodiment provides a method for collecting and detecting tunnel crack images, which is applied to the tunnel crack image collection and detection device described in Example 1. The method includes: S100: The rail vehicle 100 moves along the tunnel a set distance each time to an identification point.

[0043] Specifically, in step S100, the railcar 100 is first pushed or controlled to travel along the tunnel. Each time the railcar 100 travels a set distance, the position at which it stops after each movement serves as an identification point. At each identification point, the railcar 100 identifies cracks in the tunnel lining. After each movement, the distance detection mechanism 110 determines the distance traveled by the railcar 100 based on the rotation angle of the railcar 100's wheels, thereby determining the position of the crack along the tunnel's extension direction.

[0044] S200: The crack identification device 400 identifies cracks at the identification points, and sends the crack orientation to the control module after identifying the cracks.

[0045] Specifically, in step S200, each time the rail vehicle 100 moves to an identification point, the third drive assembly 420 drives the identification camera 410 to rotate to detect whether there are cracks on the tunnel lining. If the identification camera 410 does not identify a crack, the image acquisition module 200 is not rotated, and the rail vehicle 100 continues to move along the tunnel; if the identification camera 410 identifies a crack, the crack orientation can be driven according to the rotation angle of the servo motor in the third drive assembly 420, and the third drive assembly 420 sends the crack orientation to the control module, so that the control module controls the indexing support assembly 310 and the bottom rotating assembly 320.

[0046] S300 : The control module controls the steering adjustment module 300 to adjust the image acquisition module 200 to face the crack according to the crack orientation, and obtains the travel distance of the rail vehicle 100 .

[0047] Specifically, in step S300, after the control module receives the crack orientation, it first determines whether the crack is located in the left half or the right half of the tunnel based on the rotation angle of the servo motor in the third drive assembly 420, and then controls the second servo motor 530 to place the upper rotating disk 322 in the first position or the second position; then, based on the state of the proximity switch 340, it is determined whether the upper rotating disk 322 has rotated into place. When the proximity switch 340 senses the magnet 350, it is determined that the upper rotating disk 322 is placed in the first position. When the proximity switch 340 does not sense the magnet 350, it is determined that the upper rotating disk 322 is placed in the second position; then, the first servo motor 510 is controlled to align the image acquisition module 200 with the crack position; finally, the crack image is acquired through the image acquisition module 200. During the acquisition process, the operator can manually observe whether there is a crack in the area formed by the four straight lasers, and then verify the accuracy of the acquisition of the device.

[0048] When the image acquisition module 200 starts to acquire data, it indicates that there is a crack on the tunnel lining corresponding to the identification point. At this time, the travel distance of the rail vehicle 100 is obtained to determine the position of the crack in the tunnel along the tunnel extension direction.

[0049] S400: The image acquisition module 200 acquires a crack image and calculates the actual size of the crack according to the crack image, so that the control module generates complete crack information according to the actual size of the crack, the travel distance and the crack orientation.

[0050] Specifically, in step S400, after the image acquisition module 200 acquires the crack image, the crack image is stored and processed. The actual size of the crack is calculated by processing the crack image, thereby completing the acquisition of the crack information set. The crack information set includes the actual size of the crack, the position of the crack along the tunnel axis, and the orientation of the crack on the tunnel lining; after obtaining the crack information set, complete crack information is generated based on the crack information set.

[0051] Furthermore, step S400 specifically includes: S410: Acquire the crack image, perform crack target detection on the crack image, and locate a rectangular bounding box area of the crack.

[0052] Specifically, in step S410, the crack image collected by the image acquisition module 200 is first acquired, and then the position of the crack in the crack image is identified. After the position of the crack is located, the crack is surrounded by setting a boundary to form a rectangular boundary box area. Due to the complex surface environment of the tunnel, the collected crack image contains a large amount of non-crack information, and the existence of this information is not conducive to the extraction of crack features; the position of the crack in the image can be quickly located through target detection, and the extraction of crack features in the located area can reduce the influence of non-crack information in the image and improve the speed and effect of feature extraction. This embodiment uses the YOLOv8 model for crack target detection, and the detection results are as follows: Figure 9 As shown, Figure 9 The rectangular bounding box indicates the location of the crack in the image.

[0053] S420: Using an improved Panoptic FPN network to perform fine segmentation of crack pixels to obtain a segmented image.

[0054] Specifically, in step S420, cracks detected after the target is segmented within the rectangular bounding box, extracting all crack-related pixels from the image. Based on the Panoptic FPN network, this embodiment improves the tunnel crack detection algorithm by selecting a backbone network, implementing semantic segmentation FPN, and integrating multi-scale features, achieving refined crack extraction and segmentation.

[0055] Furthermore, the improved Panoptic FPN network includes: The encoder adopts the ResNeSt multi-level feature extractor; the decoder upsamples and concatenates multi-scale features through a feature pyramid network, and uses the decoder head to predict and output the segmentation results of crack pixels.

[0056] Specifically, ResNeSt can encode very rich features and is more suitable for tunnel surface textures; the decoder part performs multi-scale fusion and decoder head prediction, fusing feature maps of different scales extracted by the feature extraction network, and can utilize information at more scales to greatly improve the network's detection performance for small targets and effectively improve the recognition ability of complex diseases such as tunnel cracks. The network structure of this embodiment is as follows: Figure 10 shown.

[0057] S430: Performing a thinning process on the segmented image to generate a crack skeleton, and calculating the crack length and width based on the crack skeleton.

[0058] Furthermore, step S430 includes: S431: Obtain the crack skeleton using the Zhang-Suen thinning algorithm; S432: Calculate the sum of distances between adjacent pixels on the crack skeleton to obtain the pixel-level length of the crack, and use the local perpendicular line method to obtain the pixel-level width of the crack; S433: Combined with Zhang Zhengyou's calibration method, calculate the actual length and actual width of the crack.

[0059] Specifically, in step S431, the crack skeleton is extracted from the segmented image, such as Figure 11 The image a in the figure is the crack image after crack segmentation. In this embodiment, the Zhang_Suen parallel thinning algorithm is used to refine the cracks. Figure 11 After image processing, we can get Figure 11 The crack skeleton is shown in the image b. Define the crack skeleton element as 1 and the background pixel as 0. The matrix The values are as follows: ; Assume the number of pixels of the crack skeleton is , the length of the crack skeleton within the 8 neighborhood of a certain point on the crack skeleton is , ,but:

[0060] In the formula Indicates the first Pixel-centered 3×3 matrix, Represents the convolution operation.

[0061] When the crack skeleton When a pixel is connected to two adjacent pixels by an edge and a corner, Figure 12 As shown in , if the sum of the lengths of the two solid lines is calculated directly, the calculated value of the local crack length will be larger because the local direction of the crack skeleton is not considered. To solve this problem, The length of the dotted line in the figure is set to be consistent with the direction of the Li Feng skeleton, which is closer to the actual local crack length. The calculation formula for the crack length is:

[0062] In the formula Represents the pixel-level length of the crack skeleton. Since this calculation method is based on the distance between pixel centers, the length of the first and last pixels of the crack skeleton is not fully included. Let the length of the unincluded part be In this embodiment, .

[0063] For the calculation of crack width, this embodiment uses the local perpendicular method of the crack skeleton. First, the local perpendicular direction of each point on the crack skeleton is determined; then, the crack width of each point is calculated along the perpendicular direction. The local perpendicular direction is determined by the positional relationship between a point on the crack skeleton and its two adjacent points. After statistics on all possible adjacent points, it is found that there are 8 cases for the local perpendicular direction of a point, such as Figure 13 As shown, the angles with the horizontal direction are 0°, 26.6°, 45°, 63.4°, 90°, 116.6°, 135°, and 153.4° respectively. On this basis, the matrix is defined for: ; make: ; according to The value of The local vertical direction of the pixel, The corresponding relationship between the value of and the local vertical direction is shown in Table 1.

[0064] Table 1

[0065] Set the crack skeleton The pixel location point is O. In the crack segmentation image, find the first non-crack pixel points A and B on both sides of the crack skeleton along the perpendicular direction of point O, and calculate the distance between these two points and point O. 、 , then the calculation formula for the crack width at point O is as follows:

[0066] In the formula Indicates the first pixel-level crack width in pixels, is the compensation in Table 1. Because the algorithm calculates the distance from the skeleton point to the first non-crack pixel, the calculation result will include the width of some non-crack areas. The purpose of setting the compensation is to deduct the width of this part of the non-crack area, thereby improving the detection accuracy of the width. The calculation formula for the average crack width is:

[0067] In the formula Indicates the average pixel-level width of cracks.

[0068] To obtain the actual size of the crack, camera calibration is required to find the conversion relationship between the pixel coordinates of the image and the real-world coordinates. Based on this conversion relationship, the actual size corresponding to a pixel in the image can be obtained. Combined with the pixel-level size of the crack, the actual size of the crack can be obtained. The calculation formula is as follows:

[0069] In the formula is the actual length of the crack, is the actual average width of the crack, The actual size corresponding to one pixel.

[0070] After obtaining the actual size of the crack, the crack's position within the tunnel is calculated. This information primarily includes its circumferential position and its along-line position. In step S300 , the angle of the image acquisition module 200 (i.e., the circumferential position of the crack) is calculated based on the rotation angle of the servo motor of the third drive assembly 420 . The distance traveled by the railcar 100 (i.e., the along-line position of the crack) is calculated based on the distance detection mechanism 101 .

[0071] The above description is merely a preferred embodiment of the present invention and an illustration of the technical principles employed. Those skilled in the art should understand that the scope of the invention is not limited to the technical solutions formed by the specific combination of the above-mentioned technical features, but also encompasses other technical solutions formed by any combination of the above-mentioned technical features or their equivalents without departing from the inventive concept. For example, a technical solution formed by replacing the above-mentioned features with (but not limited to) technical features with similar functions disclosed in the present invention.

Claims

1. A tunnel crack image acquisition and detection device, characterized in that: include: A railcar (100), wherein the railcar (100) is provided with a distance detection mechanism (110), and the distance detection mechanism (110) is used to detect the travel distance of the railcar (100); An image acquisition device, the image acquisition device comprising an image acquisition module (200) and a steering adjustment module (300), the image acquisition module (200) being mounted on the top of the rail vehicle (100) via the steering adjustment module (300), the steering adjustment module (300) being used to drive the image acquisition module (200) to steer so as to adjust the orientation of the image acquisition module (200); A crack identification device (400), the crack identification device (400) being arranged at the front end of the image acquisition module (200), the crack identification device (400) being used to identify cracks and determine crack orientations; A control module is provided, the control module being electrically connected to the steering adjustment module (300), the crack identification device (400) and the distance detection mechanism (110), and the control module being configured to control the steering adjustment module (300) to drive the image acquisition module to steer after a crack is identified, so that the image acquisition module (200) is aligned with the crack position.

2. The tunnel crack image acquisition and detection device according to claim 1, characterized in that: The steering adjustment module (300) comprises: a graduation support assembly (310), the graduation support assembly (310) being connected to the image acquisition module (200), the graduation support assembly (310) comprising a first rotation axis (311), the first rotation axis (311) being parallel to the top surface of the rail vehicle (100); the graduation support assembly (310) being used to drive the image acquisition module (200) to rotate about the first rotation axis (311); A bottom rotating assembly (320) is provided, wherein the bottom rotating assembly (320) is connected to the rail vehicle (100) and the indexing support assembly (310), and the bottom rotating assembly (320) includes a second rotating shaft (321), wherein the second rotating shaft (321) is perpendicular to the top surface of the rail vehicle (100); and the bottom rotating assembly (320) is used to drive the indexing support assembly (310) to rotate around the second rotating shaft (321).

3. The tunnel crack image acquisition and detection device according to claim 2, characterized in that: The indexing support assembly (310) further includes: a pair of support plates (312), the pair of support plates (312) being arranged along the extension direction of the first rotating shaft (311) and mounted on the indexing support assembly (310), the support plates (312) being fan-shaped and having an arc-shaped first notch (313), the inner edge of the first notch (313) being a V-shaped guide rail; A pair of connecting members (315), wherein the pair of connecting members (315) are arranged between the pair of supporting plates (312), one end of the connecting member (315) is fixed to the first rotating shaft (311), and the other end of the connecting member (315) is fixed to the image acquisition module (200); a guide plate (316) is provided on the connecting member (315), and a plurality of V-shaped wheels (317) are rotatably mounted on the guide plate (316), and the V-shaped wheels (317) are mounted on the V-shaped guide rails on corresponding sides; A first drive assembly is in transmission connection with the first rotating shaft (311), the first drive assembly is mounted on the support plate (312) and is electrically connected to the control module, and is used to drive the first rotating shaft (311) to rotate.

4. The tunnel crack image acquisition and detection device according to claim 3, characterized in that: The bottom rotating assembly (320) further includes: an upper rotating disk (322), the upper rotating disk (322) being connected to the indexing support assembly (310) via an upper bracket (323); a lower rotating disk (324), the lower rotating disk (324) being mounted on the top of the rail vehicle (100) via a lower bracket (325), the lower rotating disk (324) and the upper rotating disk (322) being relatively rotatable around the second rotating axis (321), so that the upper rotating disk (322) can be switched between a first station and a second station, the first station and the second station having an angle difference of 180°; a locking assembly, the locking assembly being mounted on the upper rotating disk (322) and the lower rotating disk (324), respectively, the locking assembly being used to lock the upper rotating disk (322) at the first station or the second station; A second drive assembly is in transmission connection with the second rotating shaft (321), the second drive assembly is mounted on the lower rotating disk (324) and is electrically connected to the control module, and is used to drive the second rotating shaft (321) to rotate.

5. The tunnel crack image acquisition and detection device according to claim 4, characterized in that: The image acquisition module (200) comprises: An arc-shaped housing (210), the arc-shaped housing (210) being connected to the indexing support assembly (310), a light source (220) and an area array camera (230) being provided on the top of the arc-shaped housing (210), the area array camera (230) being used to collect images of the cracks; A plurality of laser infrared emitters (240) are evenly distributed around the arc-shaped housing (210), and the laser infrared emitters (240) are connected to the arc-shaped housing (210) via a damping shaft.

6. The tunnel crack image acquisition and detection device according to claim 5, characterized in that: The crack identification device (400) comprises: an identification camera (410), the identification camera (410) being rotatably mounted on the support plate (312) near the front end of the rail vehicle (100), with the rotation axis being coaxial with the first rotation axis (311); the identification camera (410) being used to identify cracks, and the identification camera (410) being electrically connected to the control module; A third driving component (420), the second driving component and the recognition camera (410) are installed on the same support plate (312), the output end of the third driving component (420) is transmission-connected to the recognition camera (410), and the third driving component (420) is used to drive the recognition camera (410) to rotate.

7. A tunnel crack image acquisition and detection method, characterized in that: Applied to the tunnel crack image acquisition and detection device according to any one of claims 1 to 6, the method comprises: The rail vehicle (100) moves along the tunnel a set distance each time to an identification point; The crack identification device (400) identifies the crack at the identification point, and sends the crack orientation to the control module after identifying the crack; The control module controls the steering adjustment module (300) to adjust the image acquisition module (200) toward the crack according to the crack orientation, and obtains the travel distance of the rail vehicle (100); The image acquisition module (200) acquires crack images and calculates the actual size of the crack based on the crack images, so that the control module generates complete crack information based on the actual size of the crack, the travel distance and the crack orientation.

8. The tunnel crack image acquisition and detection method according to claim 7, characterized in that: The image acquisition module (200) acquires crack images and calculates the actual size of the cracks based on the crack images, including: Acquire the crack image, perform crack target detection on the crack image, and locate a rectangular bounding box area of the crack; An improved Panoptic FPN network is used to perform fine segmentation of crack pixels to obtain a segmented image; The segmented image is thinned to generate a crack skeleton, and the crack length and width are calculated based on the crack skeleton.

9. The tunnel crack image acquisition and detection method according to claim 8, characterized in that: The improved Panoptic FPN network includes: The encoder adopts the ResNeSt multi-level feature extractor; the decoder upsamples and concatenates multi-scale features through a feature pyramid network, and uses the decoder head to predict and output the segmentation results of crack pixels.

10. The tunnel crack image acquisition and detection method according to claim 9, characterized in that: The step of performing thinning processing on the segmented image to generate a crack skeleton, and calculating the crack length and width based on the crack skeleton, comprises: The crack skeleton is obtained using the Zhang-Suen thinning algorithm; Calculating the sum of the distances between adjacent pixels on the crack skeleton to obtain the pixel-level length of the crack, and using the local perpendicular line method to obtain the pixel-level width of the crack; Combined with Zhang Zhengyou's calibration method, the actual length and width of the crack are calculated.

Images