Vehicle-mounted point cloud density enhancing device and method for tunnel lining disease detection

Abstract

The invention discloses a vehicle-mounted point cloud density enhancing device and method for tunnel lining disease detection, the vehicle-mounted point cloud density enhancing device is composed of an image acquisition assembly, a data transmission assembly, an intelligent terminal and a laser scanner, the image acquisition assembly and the laser scanner are integrally located at the top of a detection vehicle, the data transmission assembly is used for realizing data link, and the intelligent terminal is connected with the image acquisition assembly. And transmitting picture information obtained by the image acquisition assembly and laser scanning data to the intelligent terminal through Bluetooth. The laser scanner performs 360-degree rotary scanning on the tunnel section in the detection process to obtain a standard distance and further obtain three-dimensional information of a point, and three-dimensional point cloud data is generated by converting a coordinate system. According to the invention, the area-array camera and the grating projector are integrated, the point cloud image is generated by using the visual reconstruction technology, the laser scanner is assisted, the point cloud density of the collected data is improved, and the comprehensiveness and accuracy of disease detection are ensured.

Description

Technical Field

[0001] The present invention relates to a vehicle-mounted point cloud density enhancement device and method for tunnel lining defect detection. The device addresses the problem of overly sparse point clouds in current laser scanner applications in tunnel defect detection, which can lead to missed defects. The device encrypts the collected point cloud data to achieve efficient and comprehensive detection of defects in high-speed rail tunnels and other applications. Background Art

[0002] During operation, tunnels for high-speed railways, subways, highways, and other projects may experience defects such as lining cracks, water leakage, block falling, and geometric deformation. These defects need to be detected and treated in a timely manner to avoid affecting the normal operation of high-speed railways.

[0003] Taking high-speed rail tunnels as an example, defect detection plays a vital role in ensuring the safe and normal operation of high-speed rail. Due to the long mileage of high-speed rail tunnels and the short operating window period of high-speed rail tunnels, manual inspection methods can no longer meet the existing detection needs. At present, detection technologies based on multiple sensors and data processing methods based on computer vision are widely used in the field of tunnel defect detection, which has led to the development of tunnel defect detection technology in the direction of automation and intelligence. However, in order to meet the demand for rapid detection, the existing tunnel defect detection devices based on laser scanners rotate the scanning structure and advance the device simultaneously during the detection process, and collect data in a spiral manner in the tunnel. However, the data collected by this type of device has the problem of sparse point clouds, which will cause some defects to be missed during the detection process.

[0004] CN106886980A discloses a method for enhancing point cloud density based on three-dimensional laser radar target recognition. The method comprises measuring the initial point cloud data of the target using a three-dimensional laser radar, determining the target bounding volume in the initial point cloud data, constructing a local coordinate system with the center of the target bounding volume as the origin, converting the initial point cloud data from the initial radar coordinate system to the local coordinate system to obtain converted point cloud data, constructing a three-dimensional surface based on a radial interpolation function (RBF) and the converted point cloud data, and resampling the point cloud based on the three-dimensional surface to generate a new point cloud. The present invention can make the target point cloud density independent of its distance, which is beneficial for point cloud feature extraction. It can also directly utilize existing point cloud feature research results, such as histograms, and is applicable to any segmented target. However, this method requires establishing a local coordinate system and converting the point cloud data, and also requires data resampling, resulting in a complex process.

[0005] CN106023319A discloses a method for repairing the structural features of ground objects based on laser point clouds using CCD photos. The method first generates a projection distance image with a similar shooting angle to the CCD photo based on existing point cloud data, and records the conversion relationship between the point cloud and the projection distance image. The projection distance image and the CCD photo taken later are then image matched to obtain the spatial relative relationship between the two images. After image matching is complete, a reference point is selected on the same plane as the position to be repaired in the projection distance image, and the position to be repaired is found on the retaken CCD photo. The distance to the position to be repaired is calculated based on the spatial relative relationship of the images. Furthermore, the actual point cloud coordinates of the position to be repaired are calculated based on the conversion relationship between the point cloud and the projection distance image. Finally, based on the restored feature points, point cloud point, line, and surface feature repair is achieved through line and surface encryption of the point cloud data. This method requires post-production retakes, making it difficult to meet the requirements for completing tunnel disease detection in a short period of time.

[0006] CN109902425A discloses a method for extracting tunnel sections from ground-based point clouds, comprising: (1) calculating the center coordinates of the extracted section location; (2) determining the tunnel section equation; (3) determining the ray equation within the tunnel section; (4) determining the section points; and (5) section fitting and resampling. The present invention employs a cylindrical surface that better fits the tunnel surface shape for local fitting, thereby overcoming the influence of tunnel surface noise, being able to adapt to different tunnel point cloud densities, and being able to smoothly extract tunnel sections even for non-uniform tunnel point clouds; at the same time, further fitting and resampling processing is performed on the extracted section points to further adapt to the non-homogeneity of the point cloud and to compensate for some missing point clouds to a certain extent. However, the device has a high component cost and complex data processing.

[0007] CN117974887A discloses a method and system for modeling tunnel walls based on three-dimensional laser point clouds. The method comprises: obtaining three-dimensional laser point cloud data for tunnel wall modeling, preprocessing the three-dimensional laser point cloud data, wherein the preprocessing includes completing, filtering, simplifying, and smoothing the three-dimensional laser point cloud data; and reconstructing the preprocessed three-dimensional laser point cloud data based on a KD tree and grid method. The present invention proposes a method for completing missing point cloud data based on an improved minimum angle completion algorithm, thereby improving the quality of the final reconstructed tunnel wall triangulation model and achieving good adaptability when processing point clouds with different scales and densities. However, the data preprocessing of this method includes multiple processes such as completion, elimination, simplification, and smoothing, which can easily cause data loss. If this technology is applied to the field of tunnel disease detection, it may cause image distortion.

[0008] CN115061115A discloses a point cloud encryption method, device, storage medium, and laser radar. The method includes: obtaining the point cloud encryption multiples for each level of detection field of view; obtaining the scan line interval corresponding to two adjacent transmissions based on the point cloud encryption multiples for each level of detection field of view; and scanning according to the scan line interval corresponding to the two adjacent transmissions. Using this application, while ensuring eye safety, ROI encryption can be achieved by controlling the scan line interval corresponding to two adjacent transmission groups, thereby improving the efficiency of laser radar detection. However, the device is complex to operate, the data processing process is cumbersome, and the cost is relatively high.

[0009] CN110109127A discloses a device and method for increasing the density of laser radar point clouds. A plane mirror is added to the device, so that the reflection matrix can be obtained during calculation. In the present invention, the normal vector of the calibration plate plane and the direction vector of the edge line are used as features, and the rotation matrix is calibrated by SVD decomposition. The intersection of the edge lines of the calibration plate, that is, the coordinates of the corner points of the calibration plate, is used to calibrate the translation matrix. The result of the external parameter calibration is (R|t). The calibration result can be used to map the point cloud of the non-interest area to the interest area using a calculation formula, effectively increasing the density of the laser radar point cloud. This method requires the division of the interest area and the non-interest area. If it is applied to the field of tunnel disease detection, the detection efficiency will be reduced, and the device installation is more complicated.

[0010] Overall, these technologies have made some progress in increasing point cloud density. However, they generally suffer from complex operations, high costs, and cumbersome processes, making it difficult to achieve efficient tunnel defect detection in a short period of time. These shortcomings limit the widespread adoption and practical application of existing technologies. Therefore, there is an urgent need for a new technical solution that can simplify the operational process, reduce costs, and achieve efficient and reliable tunnel defect detection within a limited timeframe.

[0011] In summary, the current point cloud density enhancement devices / methods in the field of tunnel inspection have the following main shortcomings:

[0012] (1) The high complexity of device operation leads to low efficiency in tunnel disease detection. The complex operation steps not only increase the risk of misoperation, but also affect the accuracy and stability of detection;

[0013] (2) The device has a complex structure, which increases manufacturing and maintenance costs and makes it inconvenient to use and install. The complex mechanical structure and electronic system increase the failure rate and maintenance difficulty of the device;

[0014] (3) In order to increase the point cloud density, existing devices usually require multiple inspections, which not only prolongs the inspection time, but also makes it difficult to complete the inspection work within the limited window time of the tunnel, and cannot meet the needs of rapid inspection in high-speed railway tunnels. Summary of the Invention

[0015] The present invention aims to address the deficiencies of the above-mentioned prior art and provide an efficient, fast, reliable, and comprehensive device for tunnel lining defect detection and a corresponding method. The present invention integrates an area array camera and a grating projector into the detection device, utilizes visual reconstruction technology to generate point cloud images, and assists the laser scanner to increase the point cloud density of the collected data, thereby ensuring the comprehensiveness and accuracy of defect detection. The device can be conveniently installed on an inspection vehicle to perform 360° all-round defect detection on tunnel sections. The efficient data processing module can quickly receive and process collected images and point cloud data, perform multi-perspective data fusion, point cloud optimization, and export in a standard format to ensure data integrity and accuracy.

[0016] The technical solution of the present invention is:

[0017] A vehicle-mounted point cloud density enhancement device for tunnel lining defect detection primarily consists of an image acquisition component 1, a data transmission component 2, an intelligent terminal 3, and a laser scanner 4. The image acquisition component 1 and laser scanner 4 are located on top of the inspection vehicle. The data transmission component 2 establishes data links with the image acquisition component 1, the laser scanner 4, and the intelligent terminal 3 via Bluetooth. The image information obtained by the image acquisition component 1 and the data from the laser scanner 4 are transmitted to the intelligent terminal 3 via Bluetooth. During the inspection process, the laser scanner 4 rotates 360° across the tunnel cross-section to obtain standard distances and three-dimensional information about the points. The laser scanner then converts the coordinate system to generate three-dimensional point cloud data.

[0018] The image acquisition component 1 comprises a set of area array cameras 12 and a set of grating projectors 13. Its primary function is to compensate for data not captured by the laser scanner 4 during tunnel operation, ensuring the device's normal operation within the tunnel and the collection of high-precision tunnel defect information. Data is transmitted to the intelligent terminal 3 via the built-in Bluetooth of the area array cameras 12 and the laser scanner 4. The intelligent terminal 3 comprises a control module and a data processing module. The control module issues commands to the device, adjusting the acquisition frequency of the area array camera 12, the frequency of the laser scanner 4, and the driving speed. The data processing module, comprising a computing unit, a high-efficiency storage unit, and specialized software, converts the image data captured by the area array camera 12 into three-dimensional point cloud data, which is then displayed in a three-dimensional visualization along with the point cloud data from the laser scanner 4, enabling efficient and high-precision detection of tunnel lining defects.

[0019] The image acquisition assembly 1 consists of a vehicle-mounted mounting bracket 11, an array of area array cameras 12, and an array of grating projectors 13. The bottom of the vehicle-mounted mounting bracket 11 is a rectangular frame structure, atop which is mounted an L-shaped bracket for detecting the vehicle's direction of travel. A circular disk is located at the end of the L-shaped bracket, which is aligned with the tunnel's radial direction. The array of area array cameras 12 and the array of grating projectors 13 are mounted along the circumference of the disk (the area array cameras 12 and the grating projectors 13 are symmetrically arranged side by side around the disk's circumference). The laser scanner 4 is mounted at the center of the disk's outer surface.

[0020] Preferably, the array of area array cameras 12 comprises eight high-resolution area array cameras evenly distributed along the circumference, arranged in a circular pattern and evenly spaced on a disc mounted on a vehicle-mounted mount facing the direction of travel. Each area array camera 12 is equipped with a high-resolution sensor capable of capturing fine details of the tunnel lining surface within its viewing sector. The even distribution of the eight area array cameras ensures full coverage of the tunnel cross-section, ensuring that no areas are missed.

[0021] Preferably, the grating projector array 13 comprises eight grating projectors, located in close proximity to the area array camera array 12. The function of the grating projector array 13 is to project a Gray-coded grating onto the tunnel surface, providing reference information for the area array camera array 12 to capture images. The grating projector array 13 adopts a flat box shape, which effectively reduces the size and weight of the device and facilitates side-by-side installation with the area array camera array 12, maintaining a good relative position.

[0022] Preferably, the vehicle-mounted fixing bracket 11 is made of metal, which has excellent stability and durability, and can effectively fix and support the area array camera array and the grating projector array, ensuring that they maintain a stable and precise position during the detection process.

[0023] Preferably, the area array camera 12 array adopts a rectangular parallelepiped shape with a compact internal structure, which can effectively reduce the volume and weight of the equipment and facilitate installation and transportation.

[0024] The main function of the data transmission component is to store and package the laser scanning data and image information collected by the device, and transmit them to the smart terminal via Bluetooth. It is also responsible for transmitting the control instructions of the smart terminal to the device.

[0025] The intelligent terminal consists of a control module and a data processing module, and its main functions are as follows:

[0026] (1) Control the coordinated operation of various systems in the entire device in a complex tunnel environment, especially the frame rate and accuracy of the laser scanner and image acquisition components according to the collected information;

[0027] (2) The received image information and laser point cloud data are processed and converted into visual three-dimensional point cloud data.

[0028] Furthermore, the control module is used to adjust the camera shooting frequency to match the laser scanner frequency and the driving speed.

[0029] Furthermore, the data processing module processes the received image information and laser point cloud data, converts the image information into 3D point cloud data, and displays it in a 3D visualization together with the laser scanner point cloud data, thus achieving high-efficiency and high-precision detection of tunnel lining defects. The data processing process includes:

[0030] First, a calibration image is captured using an area array camera, and the Roger Tsai two-step calibration method is used to determine the extrinsic and intrinsic parameters. Specifically, the first step uses the least squares method to solve for the extrinsic parameters; the second step uses a three-variable optimization search to solve for the internal parameters, including the focal length f, the scale factor s, and the distortion coefficient k.

[0031] Then, the image is binarized using the Gray code grating diagram and the coding table to determine the phase information, and the fringe image is thinned to obtain the coordinates of the pixel feature points;

[0032] Finally, these coordinate information are converted into three-dimensional point cloud data, and with the help of 3DMAX modeling tools, they are imported into the tunnel lining disease detection device for analysis and processing.

[0033] A vehicle-mounted point cloud density enhancement method for tunnel lining defect detection includes the following steps:

[0034] (1) Use the vehicle-mounted bracket to install the detection device on the top of the inspection vehicle. Adjust the angles of the area array camera and projector according to the cross-sectional dimensions of the tunnel to ensure that no area is missed. Turn on the laser scanner, area array camera, and grating projector and complete the parameter initialization settings.

[0035] (2) Image calibration is performed before the detection begins. At the detection starting position, the laser scanner scans the tunnel section 360° to obtain the standard distance. At the same time, the image acquisition component captures the image, and the actual distance corresponding to the modulated grating in the image can be known;

[0036] (3) The detection vehicle's speed is matched with the frame rate of the area array camera and the frame rate of the laser scanner, maintaining a constant speed throughout the entire process, and collecting information along the track along the entire tunnel;

[0037] (4) The image data and laser point cloud data collected at a certain distance are packaged and transmitted to the intelligent terminal through the data transmission component, and the data processing module performs data processing and generates a three-dimensional point cloud;

[0038] (5) The data processing module denoises the collected laser scanning data, performs coordinate transformation to generate a point cloud, calibrates and solves the collected image information, and converts it into point cloud data. The ICP algorithm is then used to fuse the two point cloud data from multiple perspectives.

[0039] Furthermore, the ICP algorithm includes:

[0040] (a) The point cloud data converted from image data is the source point cloud, and the point cloud converted from laser scanning data is the target data. First, nearest neighbor matching is performed to find the closest point in the target point cloud for each point in the source point cloud;

[0041] (b) Then the rotation matrix R and translation vector T are obtained through transformation calculation;

[0042] (c) Then update the source point cloud. If the error between the new source point cloud and the target point cloud has not converged, perform error calculation and iteration until the error converges.

[0043] (d) Output the final updated source point cloud, i.e. the final registration result. After point cloud denoising, the processed point cloud data is exported to a standard format file (such as PLY, LAS, XYZ, etc.).

[0044] The working principle of the present invention is:

[0045] Image calibration is performed before inspection begins. A laser scanner scans the tunnel cross-section 360° to obtain a standard distance. Simultaneously, the image acquisition component captures images, revealing the actual distances corresponding to the modulated gratings in the image. The image acquisition component consists of a set of eight area array cameras and a set of eight grating projectors. The area array cameras are arranged in a circular pattern, adapting to tunnels of various cross-sectional types. During operation, the grating projectors project a Gray-coded grating onto the tunnel surface. The grating is modulated by the lining surface and captured by the area array cameras. After acquisition, the image data is transmitted to the data processing module. After a series of phase deconvolution operations, the phases are converted into a three-dimensional point cloud based on the system's calibrated parameters. This method can obtain high-precision three-dimensional data and provides good measurement accuracy and reliability within a certain range.

[0046] Beneficial effects of the present invention:

[0047] (1) The combination of surface structured light 3D measurement technology and laser scanning technology significantly improves the density of point cloud data during high-speed acquisition, reduces the omission of defects, and improves the comprehensiveness and reliability of detection results;

[0048] (2) Design a simplified device structure and automated detection process to reduce operational difficulty, reduce manual involvement, and improve the ease of use and practicality of the system;

[0049] (3) The efficient data processing module can quickly process and optimize the collected point cloud data, ensuring high accuracy and quality of the data, and providing reliable support for subsequent analysis and decision-making;

[0050] (4) The control module can monitor and adjust the operating status of the detection device in real time, ensure the continuity and stability of the detection process, and improve the real-time performance and accuracy of the detection;

[0051] (5) The present invention can effectively improve the point cloud density of tunnel lining disease detection, improve the comprehensiveness and accuracy of the detection results, and provide strong guarantees for the safe operation of the tunnel. BRIEF DESCRIPTION OF THE DRAWINGS

[0052] Figure 1 : Schematic diagram of the composition of the system of the present invention.

[0053] Figure 2 : Schematic diagram of the structure of the vehicle-mounted point cloud density enhancement device of the present invention.

[0054] Figure 3 :Working principle of image acquisition component.

[0055] Figure 4 : Schematic diagram of image calibration before detection.

[0056] Figure 5 : Data processing flow chart of monocular measurement system.

[0057] Figure 6 : Laser scanner data processing flow chart.

[0058] Figure 7 : Overall flow chart of the vehicle-mounted point cloud density enhancement method of the present invention.

[0059] Figure 8 : Schematic diagram of ICP registration algorithm.

[0060] The reference numerals in the figure are: 1-image acquisition component, 11-vehicle fixed bracket, 12-area array camera, 13-projector; 2-data transmission component; 3-intelligent terminal; 4-laser scanner. DETAILED DESCRIPTION

[0061] Example 1

[0062] like Figure 1As shown in the figure, the present invention comprises a vehicle-mounted point cloud density enhancement device for tunnel lining defect detection. The device primarily comprises an image acquisition component 1, a data transmission component 2, an intelligent terminal 3, and a laser scanner 4. The image acquisition component 1 and the laser scanner 4 are integrally located atop the inspection vehicle. The data transmission component 2 establishes data links with the image acquisition component 1, the laser scanner 4, and the intelligent terminal 3 via Bluetooth. The image information obtained by the image acquisition component 1 and the data from the laser scanner 4 are transmitted to the intelligent terminal 3 via Bluetooth. During the inspection process, the laser scanner 4 rotates 360° across the tunnel cross-section to obtain standard distances and thus acquire three-dimensional information about the points. The laser scanner 4 then generates three-dimensional point cloud data by converting the coordinate system.

[0063] The image acquisition component 1 comprises a set of area array cameras 12 and a set of grating projectors 13. Its primary function is to compensate for data not captured by the laser scanner 4 during tunnel operation, ensuring the device's normal operation within the tunnel and the collection of high-precision tunnel defect information. Data is transmitted to the intelligent terminal 3 via the built-in Bluetooth of the area array cameras 12 and the laser scanner 4. The intelligent terminal 3 comprises a control module and a data processing module. The control module issues commands to the device, adjusting the acquisition frequency of the area array camera 12, the frequency of the laser scanner 4, and the driving speed. The data processing module, comprising a computing unit, a high-efficiency storage unit, and specialized software, converts the image data captured by the area array camera 12 into three-dimensional point cloud data, which is then displayed in a three-dimensional visualization along with the point cloud data from the laser scanner 4, enabling efficient and high-precision detection of tunnel lining defects.

[0064] like Figure 2 As shown, the image acquisition assembly 1 comprises a vehicle-mounted mounting bracket 11, an array of area array cameras 12, and an array of grating projectors 13. The bottom of the vehicle-mounted mounting bracket 11 is a rectangular frame structure, atop which is mounted an L-shaped bracket for detecting the direction of vehicle travel. A circular disk is located at the end of the L-shaped bracket, which is aligned with the radial direction of the tunnel. The array of area array cameras 12 and the array of grating projectors 13 are symmetrically arranged side by side on the disk, and the laser scanner 4 is mounted at the center of the outer side of the disk.

[0065] Preferably, the array of area array cameras 12 comprises eight high-resolution area array cameras evenly distributed along the circumference, arranged in a circular pattern and evenly spaced on a disc mounted on a vehicle-mounted mount facing the direction of travel. Each area array camera 12 is equipped with a high-resolution sensor capable of capturing fine details of the tunnel lining surface within its viewing sector. The even distribution of the eight area array cameras ensures full coverage of the tunnel cross-section, ensuring that no areas are missed.

[0066] Preferably, the grating projector array 13 comprises eight grating projectors, located in close proximity to the area array camera array 12. The function of the grating projector array 13 is to project a Gray-coded grating onto the tunnel surface, providing reference information for the area array camera array 12 to capture images. The grating projector array 13 adopts a flat box shape, which effectively reduces the size and weight of the device and facilitates side-by-side installation with the area array camera array 12, maintaining a good relative position.

[0067] Preferably, the vehicle-mounted fixing bracket 11 is made of metal, which has excellent stability and durability, and can effectively fix and support the area array camera array and the grating projector array, ensuring that they maintain a stable and precise position during the detection process.

[0068] Preferably, the area array camera 12 array adopts a rectangular parallelepiped shape with a compact internal structure, which can effectively reduce the volume and weight of the equipment and facilitate installation and transportation.

[0069] The main function of the data transmission component is to store and package the laser scanning data and image information collected by the device, and transmit them to the smart terminal via Bluetooth. It is also responsible for transmitting the control instructions of the smart terminal to the device.

[0070] The intelligent terminal is the "brain" of the entire device and consists of a control module and a data processing module. Its main functions are as follows:

[0071] First, it controls the coordinated work of various systems in the entire device in a complex tunnel environment, especially coordinating and controlling the frame rate and accuracy of the laser scanner and image acquisition components based on the collected information; second, it processes the received image information and laser point cloud data into visual three-dimensional point cloud data.

[0072] The main function of the control module is to issue various instructions to the detection device and adjust the camera shooting frequency to match the laser scanner frequency and driving speed.

[0073] The data processing module is mainly composed of computer system equipment such as computing units, high-efficiency storage units, and professional software. Its functions are as follows: First, it receives packaged files of sampling images and laser scanning data; second, it generates adaptive shooting frequencies and resolutions for each inspection location in the tunnel based on the tunnel's refined sampling information, taking into account the storage and collection of effective data; third, it converts the detection data of the binocular measurement system into 3D point cloud data, and displays it in a 3D visual manner together with the laser scanner point cloud data, realizing high-efficiency and high-precision detection of tunnel lining defects.

[0074] The data processing procedures are shown in Figure 5 and Figure 6 shown.

[0075] See also Figure 5 First, a calibration image is captured using an area array camera, and the Roger Tsai two-step calibration method is used to determine the extrinsic and intrinsic parameters. Specifically, the first step uses the least squares method to solve for the extrinsic parameters; the second step uses a three-variable optimization search to solve for the internal parameters, including the focal length f, the scale factor s, and the distortion coefficient k. The image is then binarized using a Gray code grating diagram and a coding table to determine the phase information. The fringe image is then refined to obtain the coordinates of the pixel feature points. Finally, this coordinate information is converted into 3D point cloud data and imported into the tunnel lining defect detection device using the 3DMAX modeling tool for analysis and processing.

[0076] See also Figure 6 For the data acquired by the laser scanner, the noise data must be removed first, and then the denoised data must be subjected to coordinate transformation processing, that is, the point cloud data in different coordinate systems are unified into a common coordinate system for subsequent analysis and processing; finally, the point cloud is generated based on the transformed coordinate information.

[0077] After the collected data is converted into 3D point cloud data, multi-view data fusion is required, that is, the point cloud data obtained from scanning at different positions are registered and aligned. The ICP registration algorithm can be used to merge multiple point cloud data into a whole. The principle of the ICP registration algorithm is as follows: Figure 8 As shown in the figure, point cloud optimization is then performed to remove irregular noise from the point cloud surface and improve data accuracy. Finally, the processed point cloud data is exported to a standard format file (such as PLY, LAS, XYZ, etc.). The accuracy and integrity of the point cloud data can be viewed and verified using point cloud processing software.

[0078] The working principle of the device and method of this embodiment is as follows:

[0079] As shown in Figure 4, image calibration is performed before the detection begins. The laser scanner scans the tunnel section 360° to obtain the standard distance. At the same time, the image acquisition component captures the image, and the actual distance corresponding to the modulated grating in the image can be known.

[0080] See also Figure 2 and Figure 3As shown, the image acquisition component's primary function is to capture data from areas not captured by the laser scanner. It consists of a set of eight area array cameras and a set of eight grating projectors, arranged in a circular pattern to accommodate tunnels of various cross-sectional types. During system operation, the grating projectors project a Gray-coded grating onto the tunnel surface. The grating is modulated by the lining surface and captured by the area array cameras. After acquisition, the image data is transmitted to the data processing module. After a series of phase deconvolution operations, the phases are converted into a three-dimensional point cloud based on system-calibrated parameters. This method can obtain high-precision three-dimensional data and offers good measurement accuracy and reliability within a certain range.

[0081] Example 2

[0082] See also Figure 7 The vehicle-mounted point cloud density enhancement method for tunnel lining defect detection of the present invention comprises the following steps:

[0083] (1) Use the vehicle-mounted bracket to install the detection device on the top of the inspection vehicle. Adjust the angles of the area array camera and projector according to the cross-sectional dimensions of the tunnel to ensure that no areas are missed. Turn on the laser scanner, area array camera, and grating projector and complete the parameter initialization settings. Adjust the resolution of the area array camera to an accuracy of 0.2 mm for identifying tunnel cracks (for example, when the field of view width is 2 meters, the resolution needs to reach at least 10,000 pixels horizontally).

[0084] (2) See Figure 4 ,Image calibration is performed before the detection begins.,At the starting position of the detection, the laser scanner scans the tunnel section 360°,to obtain the standard distance. At the same time, the image acquisition component,collects the image, and the actual distance corresponding to the modulated,grating in the image can be known;

[0085] (3) The driving speed was set to 60 km / h, the frame rate of the area array camera was set to 83350 Hz, and the frame rate of the laser scanner was set to 200 Hz. A constant speed was maintained throughout the entire process, and information was collected along the track along the entire tunnel.

[0086] (4) See Figure 1 , the image data and laser point cloud data collected every 5m are packaged and transmitted to the intelligent terminal through the data transmission component, and the data processing module performs data processing and generates three-dimensional point clouds;

[0087] (5) The data processing module removes noise from the collected laser scanning data and then performs coordinate transformation to generate a point cloud (see Figure 6 ), calibrate and solve the collected image information and convert it into point cloud data (see Figure 5 ). Then the ICP algorithm (see Figure 8) The point cloud data of the two are fused from multiple perspectives. Specifically, the following steps are performed: ① The point cloud data converted from image data is the source point cloud, and the point cloud converted from laser scanning data is the target data. First, nearest neighbor matching is performed to find the closest point in the target point cloud for each point in the source point cloud. ② The rotation matrix R and translation vector T are then calculated through transformation. ③ The source point cloud is then updated. If the error between the new source point cloud and the target point cloud has not converged, the error calculation and iteration are performed until the error converges. ④ The final updated source point cloud is output, which is the final registration result. After point cloud denoising, the processed point cloud data is exported to a standard format file (such as PLY, LAS, XYZ, etc.).

[0088] (6) After completing a tunnel inspection, turn off the power supply of the device, clean and inspect it to ensure that there is no damage or abnormality. Place the device in the designated location in the vehicle and ensure that it is in a safe and stable state. Perform regular maintenance to ensure the normal operation and service life of the device.

[0089] Through the above embodiments, it is not difficult for those skilled in the art to obtain the following results:

[0090] (1) The present invention can effectively improve the point cloud density

[0091] A comparison revealed that the density of the fused point cloud data was significantly higher than that of point cloud data collected using only the laser scanner. The number of points increased by approximately 10 times, enabling a clearer display of the detailed features of the tunnel lining surface and effectively compensating for the sparse point cloud obtained with the laser scanner.

[0092] (2) Improved disease detection results

[0093] Using fused point cloud data for defect detection can more comprehensively identify defects such as cracks, water leakage, and falling blocks in the tunnel lining. The detection accuracy has increased by approximately 20%, effectively reducing the risk of missing defects.

Claims

1. A vehicle-mounted point cloud density enhancement device for tunnel lining defect detection, characterized in that: The invention comprises an image acquisition component (1), a data transmission component (2), an intelligent terminal (3) and a laser scanner (4), wherein the image acquisition component (1) and the laser scanner (4) are integrally located on the top of the inspection vehicle, the data transmission component (2) respectively realizes data link with the image acquisition component (1), the laser scanner (4) and the intelligent terminal (3) via Bluetooth, and transmits the picture information obtained by the image acquisition component (1) and the data of the laser scanner (4) to the intelligent terminal (3) via Bluetooth; during the inspection process, the laser scanner (4) rotates and scans the tunnel section 360 degrees to obtain the standard distance and thus obtains the three-dimensional information of the point, and generates three-dimensional point cloud data by converting the coordinate system; the image acquisition component (1) is used to make up for the area data not collected by the laser scanner (4) during the walking process; The image acquisition component (1) includes an array of several area array cameras (12) and an array of grating projectors (13) symmetrically arranged with the area array cameras (12), wherein the area array cameras (12) are used to acquire images, and the grating projectors (13) are used to project Gray-coded gratings onto the tunnel surface; the intelligent terminal (3) is used to achieve mutual matching among the shooting frequency of the area array cameras (12), the frequency of the laser scanner (4) and the driving speed, and convert the image data collected by the area array cameras (12) into three-dimensional point cloud data, which is displayed together with the point cloud data of the laser scanner (4) in a three-dimensional visual manner to achieve point cloud density enhancement.

2. The vehicle-mounted point cloud density enhancement device for tunnel lining defect detection according to claim 1 is characterized in that: The image acquisition component (1) further comprises a vehicle-mounted fixed frame (11), the bottom of the vehicle-mounted fixed frame (11) adopts a rectangular frame structure, an "L"-shaped bracket pointing to the detection direction of the vehicle is arranged on the rectangular frame structure, a disk is arranged at the end of the "L"-shaped bracket pointing to the detection direction of the vehicle, the radial direction of the disk is consistent with the radial direction of the tunnel, the array of area array cameras (12) and the array of grating projectors (13) are installed on the circumference of the disk, and the laser scanner 4 is installed at the center of the outer side of the disk.

3. The vehicle-mounted point cloud density enhancement device for tunnel lining defect detection according to claim 1 is characterized in that: The area array camera (12) and the grating projector (13) are symmetrically arranged side by side on the circumference of the disk.

4. The vehicle-mounted point cloud density enhancement device for tunnel lining defect detection according to claim 1 is characterized in that: The array of the area array cameras (12) consists of 8 high-resolution area array cameras evenly distributed along the circumference, and the array of the grating projectors (13) consists of 8 grating projectors. The 8 high-resolution area array cameras and the 8 grating projectors are evenly distributed on a disk of a vehicle-mounted fixed frame located in the direction of travel. Each area array camera is equipped with a high-resolution sensor for capturing fine details of the tunnel lining surface in its viewing angle sector.

5. The vehicle-mounted point cloud density enhancement device for tunnel lining defect detection according to claim 1 is characterized in that: The intelligent terminal (3) is composed of a control module and a data processing module, and is used to realize the following functions: a. Coordinate and control the laser scanner (4) and the image acquisition component (1) according to the acquired information, and the frame rate and accuracy of the acquisition; b. The received image information and laser point cloud data are processed and converted into visual 3D point cloud data.

6. The vehicle-mounted point cloud density enhancement device for tunnel lining defect detection according to claim 5 is characterized in that: The data processing flow of the data processing module includes: First, an area array camera is used to capture calibration images, and the Roger Tsai two-step method is used to calibrate the extrinsic and intrinsic parameters. Then, the image is binarized using the Gray code grating diagram and the coding table to determine the phase information, and the fringe image is thinned to obtain the coordinates of the pixel feature points; Finally, these coordinate information are converted into three-dimensional point cloud data, and with the help of 3DMAX modeling tools, they are imported into the tunnel lining disease detection device for analysis and processing.

7. The vehicle-mounted point cloud density enhancement device for tunnel lining defect detection according to claim 6 is characterized in that The two-step legal standard for liberating external and internal parameters includes: The first step is to use the least squares method to solve the external parameters; In the second step, a three-variable optimization search is used to solve the internal parameters, including the focal length f, the scale factor s, and the distortion coefficient k.

8. A vehicle-mounted point cloud density enhancement method for tunnel lining defect detection, characterized in that: The vehicle-mounted point cloud density enhancement device for tunnel lining defect detection according to any one of claims 1 to 7 comprises the following steps: (1) Use the vehicle-mounted bracket to install the detection device on the top of the inspection vehicle. Adjust the angles of the area array camera and projector according to the cross-sectional dimensions of the tunnel to ensure that no area is missed. Turn on the laser scanner, area array camera, and grating projector and complete the parameter initialization settings. (2) Image calibration is performed before the detection begins. At the detection starting position, the laser scanner scans the tunnel section 360° to obtain the standard distance. At the same time, the image acquisition component captures the image, and the actual distance corresponding to the modulated grating in the image can be known; (3) The detection vehicle's speed is matched with the frame rate of the area array camera and the frame rate of the laser scanner, maintaining a constant speed throughout the entire process, and collecting information along the track along the entire tunnel; (4) The image data and laser point cloud data collected at a certain distance are packaged and transmitted to the intelligent terminal through the data transmission component, and the data processing module performs data processing and generates a three-dimensional point cloud; (5) The data processing module denoises the collected laser scanning data, performs coordinate transformation to generate a point cloud, calibrates and solves the collected image information, and converts it into point cloud data. The ICP algorithm is then used to fuse the two point cloud data from multiple perspectives.

9. The vehicle-mounted point cloud density enhancement method for tunnel lining defect detection according to claim 8 is characterized in that: The ICP algorithm includes: (a) The point cloud data converted from image data is the source point cloud, and the point cloud converted from laser scanning data is the target data. Nearest neighbor matching is performed to find the closest point in the target point cloud for each point in the source point cloud. (b) Obtain the rotation matrix R and translation vector T through transformation calculation; (c) Update the source point cloud. If the error between the new source point cloud and the target point cloud has not converged, perform error calculation and iteration until the error converges. (d) Output the final updated source point cloud, that is, the final registration result. After point cloud denoising, the processed point cloud data is exported as a standard format file.

Images