Tunnel point cloud data rapid acquisition system and method

Abstract

The invention relates to a tunnel point cloud data rapid acquisition system and method. The system comprises a vehicle-mounted three-dimensional laser scanning unit, a handheld three-dimensional laser scanning unit and a processing unit. A vehicle-mounted three-dimensional laser scanning unit scans point cloud data of a three-dimensional space in the tunnel and records the position of a train, so that the point cloud data is associated with the position information, and first point cloud data is obtained; the handheld three-dimensional laser scanning unit performs supplementary scanning on a sheltered area and / or a section in the tunnel under the condition that the handheld three-dimensional laser scanning unit is held by a worker so as to obtain second point cloud data; the processing unit constructs a three-dimensional model of the tunnel based on the first point cloud data and the second point cloud data. The advantages of the vehicle-mounted three-dimensional laser scanning unit and the handheld three-dimensional laser scanning unit are combined, data collection can be carried out under the condition that the whole section of the urban rail tunnel is effectively covered, rapid and high-precision data collection can be achieved in the complex and narrow tunnel environment, and the efficiency and adaptability of a traditional measurement method are remarkably improved.

Description

Technical Field

[0001] The present invention relates to the field of urban rail transportation technology, and in particular to a system and method for quickly acquiring tunnel point cloud data. Background Art

[0002] In the digital management of urban rail infrastructure, three-dimensional models serve as the basic model carrier, and their construction efficiency and accuracy directly affect the intelligent level of operation and maintenance management. Traditional modeling methods such as manual surveying and BIM modeling have significant limitations: manual surveying relies on manual operation, which is not only time-consuming and labor-intensive, but also subject to human error, making it difficult to ensure millimeter-level accuracy; while BIM modeling is suitable for the design stage, it lacks the ability to capture the dynamic changes of complex equipment in actual operation and maintenance, resulting in delayed model updates. In addition, traditional methods require data collection and modeling to be completed in stages, which is a cumbersome process and cannot meet the large-scale, high-frequency rapid modeling needs of urban rail systems. For example, the measurement of geometric parameters and deformation monitoring of structures such as tracks and bridges often takes weeks or even months to complete, seriously restricting the real-time nature of digital management.

[0003] Laser point cloud data, with its high precision and non-contact measurement advantages, is becoming a mainstream method for rapidly constructing 3D models. By emitting a laser beam and measuring the reflected signal, laser scanning technology can obtain a high-density 3D coordinate point cloud on the target object's surface, with spatial resolution reaching millimeter levels. This technology not only avoids errors and equipment damage caused by physical contact, but also adapts to the modeling needs of complex shapes or fragile objects. For example, in urban rail applications, laser point clouds have been applied in areas such as tunnel cross-section inspection and track subsidence monitoring, providing high-fidelity spatial data support for digital twin systems.

[0004] However, laser scanning technology faces multiple technical bottlenecks in the 3D modeling of urban rail infrastructure. First, the problem of limited acquisition speed is particularly prominent. Traditional laser scanning equipment often takes a long time to complete data acquisition when facing large-scale areas (such as long-distance tunnels). Taking urban rail tunnels as an example, their complex geometric structure and narrow space will significantly reduce scanning efficiency. However, when the acquisition accuracy is improved to obtain higher-resolution data, the scanning speed will further decrease, forming a contradiction between accuracy and efficiency. This contradiction makes it difficult for existing technologies to meet the urgent needs of urban rail systems for rapid modeling.

[0005] Secondly, the lack of portability of the equipment limits the expansion of application scenarios. Current mainstream laser scanning equipment is generally large and heavy, which makes its deployment and operation in confined spaces such as urban rail tunnels significantly difficult. The movement of the equipment requires the use of dedicated transportation vehicles or manual handling, which not only increases the complexity of the operation, but may also limit the scanning angle due to space constraints, thereby affecting data integrity. This physical limitation directly restricts the flexible application of laser scanning technology in urban rail operation and maintenance scenarios, reducing the convenience and efficiency of on-site operations.

[0006] Finally, insufficient data quality stability becomes a key obstacle to modeling accuracy. In complex environments such as urban rail tunnels and underground structures, the performance of laser scanners is easily affected by factors such as line of sight and surface reflectivity differences. For example, the high reflectivity of metal components in tunnels can lead to oversaturated signals, while the low reflectivity of concrete walls can cause data loss. These factors collectively lead to noise points, voids, or geometric deviations in the acquisition results, forcing technicians to invest a lot of time in data cleaning, completion, and error correction, significantly increasing the workload of post-processing. This data quality issue not only prolongs the modeling cycle but also poses challenges to the reliability of the model.

[0007] For example, CN112731440A discloses a method and device for detecting deformation of high-speed railway slopes. The method includes: using a high-speed railway slope point cloud acquisition system to acquire three-dimensional point cloud data of slopes along the high-speed railway. The high-speed railway slope point cloud acquisition system includes: a laser radar device, a mileage positioning synchronization unit, and a point cloud data acquisition server; using an iterative closest point algorithm to perform point cloud registration processing on the three-dimensional point cloud data based on the benchmark point cloud data in the high-speed railway scene; extracting current slope point cloud data from the three-dimensional point cloud data after point cloud registration processing, and extracting benchmark slope point cloud data from the benchmark point cloud data in the high-speed railway scene; performing voxelization processing on the current slope point cloud data and the benchmark slope point cloud data respectively; and performing high-speed railway slope deformation detection based on the voxelization results. This technical solution uses an iterative closest point algorithm (ICP) for point cloud registration, but ICP has high computational complexity when processing large-scale point clouds, is sensitive to initial positions, and is prone to falling into local optimal solutions. In complex scenes such as urban rail slopes, multi-angle scanning by the equipment leads to inconsistent point cloud coordinate systems. Combined with environmental noise interference (such as vegetation occlusion and rain and snow reflection), it is easy to cause accumulation of registration errors, which directly affects the accuracy of subsequent deformation detection.

[0008] Furthermore, this technical solution discretizes the point cloud into a regular grid through voxelization to reduce the amount of data, but this process loses the geometric features of small deformations on the slope surface (such as cracks and settlement). The lidar and mileage positioning unit used in this technical solution are large in size and weight, making it difficult to flexibly deploy them in steep slope areas or narrow spaces (such as tunnel entrances). In addition, the movement of the equipment relies on manual or fixed tracks, and the scanning position cannot be adjusted quickly, resulting in data coverage blind spots or repeated collection, reducing overall efficiency.

[0009] As described above, the present invention proposes a method and system for quickly collecting tunnel point cloud data, hoping to solve the defects of the existing technology.

[0010] In addition, on the one hand, there are differences in understanding among those skilled in the art; on the other hand, the applicant studied a large number of documents and patents when making the present invention, but due to space limitations, not all details and contents are listed in detail. However, this does not mean that the present invention does not have the characteristics of these prior arts. On the contrary, the present invention already has all the characteristics of the prior art, and the applicant reserves the right to add relevant prior art to the background technology. Summary of the Invention

[0011] Laser scanning technology faces three core challenges in urban rail 3D modeling. First, it's difficult to improve acquisition efficiency. Traditional equipment takes a long time to scan large scenes like long tunnels. The narrow spaces and complex structures inherently slow down the process, while the pursuit of high precision further reduces acquisition efficiency, resulting in a trade-off between accuracy and speed. This contradiction directly impacts the realization of rapid urban rail modeling.

[0012] Secondly, the equipment lacks portability. Mainstream scanning instruments are large and heavy, making them difficult to maneuver in confined spaces like tunnels. Deployment requires transportation or manual handling, which is not only cumbersome but can also limit scanning angles due to space constraints, reducing data integrity and hampering field operation efficiency.

[0013] Finally, data quality and stability are poor. In complex environments like tunnels and underground, laser scanning is susceptible to variations in reflectivity and line-of-sight. For example, metal components can cause signal overload, while concrete walls can lead to data loss, ultimately resulting in noise, voids, or geometric distortion. This requires technicians to invest significant time in data cleaning and repair, which not only prolongs modeling cycles but also compromises model reliability.

[0014] To address the shortcomings of the existing technology, the present invention provides, in its first aspect, a rapid tunnel point cloud data acquisition system. The system comprises a vehicle-mounted 3D laser scanning unit, a handheld 3D laser scanning unit, and a processing unit. The vehicle-mounted 3D laser scanning unit scans point cloud data of the three-dimensional space within the tunnel and records the train's position, thereby correlating the point cloud data with the position information to obtain first point cloud data. The handheld 3D laser scanning unit, while held by a worker, performs supplementary scanning of obstructed areas and / or cross-sections within the tunnel to obtain second point cloud data. The processing unit constructs a 3D model of the tunnel based on the first and second point cloud data.

[0015] The present invention combines the advantages of both vehicle-mounted and handheld three-dimensional laser scanning units, and can collect data while effectively covering the entire cross-section of urban rail tunnels. It can achieve fast and high-precision data collection in complex and narrow tunnel environments, significantly improving the efficiency and adaptability of traditional measurement methods.

[0016] Specifically, this invention utilizes the collaborative operation of vehicle-mounted and handheld 3D laser scanning units to efficiently and comprehensively scan the entire cross-section of tunnels along urban rail transit lines. This method effectively addresses the difficulty faced by traditional scanning technologies in balancing efficiency and accuracy in complex tunnel environments. By covering specialized scenarios such as densely populated areas and areas with unusual structures, it ensures the integrity and accuracy of data collection. This method is particularly useful in applications such as tunnel structure modeling and equipment status monitoring, providing high-resolution spatial data support.

[0017] The system is highly adaptable and can flexibly adapt to tunnel scenarios of varying sizes, shapes, and complexity. Mobile scanning equipment relies on rail-mounted vehicles to rapidly capture data over a wide area, while handheld devices perform supplemental scanning in confined areas (such as narrow passages and equipment-intensive areas), creating a "primary scan + supplemental scan" operation mode. This combined approach maintains stable operation even under complex structures and variable obstructions, making it widely applicable to the 3D data acquisition needs of urban rail transit and underground facilities, including subway tunnels, underground pipeline corridors, and civil air defense projects.

[0018] Data splicing achieves high precision, employing the Iterative Closest Point (ICP) algorithm for rapid registration and fusion of multi-source point cloud data. By optimizing rotation matrix and translation vector parameters, coordinate deviations between different scanning positions are eliminated, significantly improving the spatial consistency of the point cloud data. The resulting 3D model accurately reproduces the spatial layout of the tunnel proper and its ancillary facilities, providing high-precision and highly stable basic data support for subsequent engineering design, structural analysis, and defect monitoring applications.

[0019] Preferably, the vehicle-mounted 3D laser scanning unit includes a first 3D laser scanner and a non-contact odometer. The first 3D laser scanner is mounted on the rail vehicle and is used to scan point cloud data of equipment and structures in the three-dimensional space inside the tunnel while the rail vehicle is in motion. The non-contact odometer is mounted on the rail vehicle and is used to record the trajectory of the rail vehicle, thereby synchronizing position information and point cloud data.

[0020] The collaborative operation of the first 3D laser scanner and the non-contact odometry enables continuous dynamic scanning of the rail vehicle during operation, covering a large area of ​​the tunnel and significantly improving data collection efficiency. The non-contact odometry records the trajectory in real time, ensuring the precise binding of point cloud data with spatial location information, providing a reliable spatiotemporal reference for subsequent data splicing and modeling.

[0021] According to a preferred embodiment, the vehicle-mounted 3D laser scanning unit further includes a track gauge mounted on the rail operation vehicle. The track gauge monitors the distance between the tracks in the tunnel in real time, allowing the processing unit to select the first point cloud data that meets track safety and facility maintenance requirements based on the distance.

[0022] The gauge meter monitors track spacing in real time, ensuring that scanned data complies with track safety standards. This prevents data deviations caused by track deformation or anomalies, improving the efficiency of facility maintenance. This invention optimizes point cloud data screening through track parameter feedback, enabling the system to flexibly adapt to tunnel scenarios with different track gauges (such as subways and railways), expanding the device's versatility and engineering adaptability.

[0023] According to a preferred embodiment, the processing unit is configured to: remove noise points from the first or second point cloud data; splice the first and second point cloud data at different locations to generate standard spliced ​​point cloud data; and construct a three-dimensional model of the tunnel based on the standard point cloud data. This invention eliminates invalid data interference by removing noise points, thereby improving the purity and reliability of the point cloud data.

[0024] According to a preferred embodiment, the step of removing noise points from the first point cloud data or the second point cloud data by the processing unit includes: calculating and verifying the parameters of the sphere based on a random sampling consistency algorithm and the first point cloud data or the second point cloud data; calculating the center coordinates and radius of the segment slice point cloud based on the conditional equation and the indirect adjustment method; calculating the deviation distance between the center coordinates of the segment slice point cloud and the center coordinates of the sphere; comparing the deviation distance with the radius of the sphere, and removing points whose difference is greater than the difference threshold to achieve noise point removal.

[0025] By calculating the sphere's parameters, the present invention can effectively distinguish noise points from valid data, avoiding the accidental rejection or omission of critical information. By comparing the deviation distance with the threshold of the sphere's radius, the present invention dynamically filters outliers, ensuring that the spatial distribution of the point cloud data conforms to the tunnel's structural characteristics, providing high-quality input for subsequent analysis.

[0026] According to a preferred embodiment, the step of the processing unit stitching the first point cloud data and the second point cloud data at different locations includes: calibrating the first point cloud data and the second point cloud data based on a known calibration target; performing coordinate system conversion on the calibrated first point cloud data and the second point cloud data to obtain third point cloud data and fourth point cloud data in a global coordinate system; aligning the third point cloud data and the fourth point cloud data at different locations based on position information to achieve registration; and minimizing the geometric distance between the two point clouds based on an iterative nearest point algorithm.

[0027] Through target calibration and coordinate system conversion, systematic errors between different scanning positions are eliminated, ensuring seamless splicing of the first and second point cloud data in a unified coordinate system. Based on the ICP algorithm, this method minimizes the geometric distance between point clouds, achieving millimeter-level precision in point cloud alignment, significantly improving the spatial continuity and detail restoration capabilities of 3D models.

[0028] According to a preferred embodiment, the step of the processing unit performing data splicing on the first point cloud data and the second point cloud data at different positions further includes: adjusting the standard point cloud data based on a global optimization algorithm; and correcting splicing points with large errors in the standard point cloud data.

[0029] This method uses a global optimization algorithm to adjust local errors in the spliced ​​point cloud, eliminating outliers introduced by registration bias or noise, and further improving the overall accuracy of the model. The corrected standard point cloud data maintains consistency during long-term monitoring and multi-period comparisons, providing reliable time series data support for applications such as tunnel deformation analysis and disease monitoring.

[0030] From a second aspect, the present invention provides a method for quickly collecting tunnel point cloud data, which includes: using a vehicle-mounted three-dimensional laser scanning unit to scan the point cloud data of the three-dimensional space inside the tunnel and record the position of the train, so that the point cloud data is associated with the position information to obtain first point cloud data; when the worker holds the handheld three-dimensional laser scanning unit, the handheld three-dimensional laser scanning unit is used to perform a supplementary scan of the blocked area and / or cross-section in the tunnel to obtain second point cloud data; and the processing unit constructs a three-dimensional model of the tunnel based on the first point cloud data and the second point cloud data.

[0031] The vehicle-mounted three-dimensional laser scanning unit realizes continuous dynamic scanning through the movement of the track operation vehicle, and compared with the traditional static scanning (such as point-by-point measurement of a total station), the data acquisition time is greatly shortened. The application performs supplementary scanning on the shielding area and the section, and can make up for the details missing in dynamic scanning. Through the cooperative operation of vehicle-mounted scanning and handheld scanning, and in combination with dynamic position information recording, the application realizes efficient and accurate acquisition of tunnel point cloud data.

[0032] According to a preferred embodiment, the method further comprises: removing noise points from the first point cloud data or the second point cloud data; splicing the first point cloud data and the second point cloud data at different positions to obtain spliced standard point cloud data; and constructing a three-dimensional model of the tunnel based on the standard point cloud data.

[0033] The application can reduce the amount of redundant data and improve the efficiency of subsequent data processing by quickly removing invalid or abnormal points (such as outliers caused by environmental interference and device jitter during scanning), thereby indirectly speeding up the overall scanning process. The data splicing processing method of the application can realize rapid alignment and fusion, and reduce the need for repeated scanning.

[0034] According to a preferred embodiment, the method further comprises: adjusting the standard point cloud data based on a global optimization algorithm; and correcting the splicing points with large errors in the standard point cloud data.

[0035] The application can reduce the need for repeated scanning caused by error accumulation by correcting systematic errors in the point cloud data through an iterative optimization algorithm. For local errors (such as mispositioning and overlapping) in the splicing area, the application can quickly correct the errors through local recalibration or interpolation algorithms, thereby improving data consistency and reducing rework caused by data quality problems. BRIEF DESCRIPTION OF DRAWINGS

[0036] Figure 1 is a simplified module connection relationship diagram of a tunnel point cloud data rapid acquisition system provided by the application;

[0037] Figure 2 is a working scene diagram of a vehicle-mounted three-dimensional laser scanning unit provided by the application;

[0038] Figure 3 is a working scene diagram of a handheld three-dimensional laser scanning unit in a tunnel provided by the application;

[0039] Figure 4 is a display interface diagram of a terminal provided by the application;

[0040] Figure 5 is a diagram of scanning a tunnel using a handheld three-dimensional laser scanning unit provided by the application;

[0041] Figure 6is a schematic view of one angle of the tunnel inner pipe piece provided by the present application;

[0042] Figure 7 is a schematic view of another angle of the tunnel inner pipe piece provided by the present application;

[0043] Figure 8 is a schematic view of the tunnel inner rail structure provided by the present application;

[0044] Figure 9 is a schematic view of the display interface of data processing provided by the present application;

[0045] Figure 10 is a schematic view of the display interface of data import provided by the present application;

[0046] Figure 11 is a schematic view of the display interface of data splicing provided by the present application;

[0047] Figure 12 is an enlarged schematic view of the display interface of data splicing provided by the present application

[0048] Figure 13 is a first schematic view of the display interface of data processing provided by the present application;

[0049] Figure 14 is a second schematic view of the display interface of data processing provided by the present application;

[0050] Figure 15 is a third schematic view of the display interface of data processing provided by the present application;

[0051] Figure 16 is a schematic view of the data processing effect provided by the present application;

[0052] Figure 17 is another schematic view of the data processing effect provided by the present application;

[0053] Figure 18 is a first schematic view of the aerial angle of the urban rail transit provided by the present application;

[0054] Figure 19 is a second schematic view of the aerial angle of the urban rail transit provided by the present application;

[0055] Figure 20 is a third schematic view of the aerial angle of the urban rail transit provided by the present application;

[0056] Figure 21 is Figure 2 is a working scene photo of the vehicle-mounted three-dimensional laser scanning unit shown in the figure;

[0057] Figure 22 is Figure 3 The photo of the handheld 3D laser scanning unit in operation is shown;

[0058] Figure 23 yes Figure 4 An enlarged schematic diagram of the display interface of the terminal is shown.

[0059] Reference Signs List

[0060] 100: Track operation vehicle; 200: Vehicle-mounted 3D laser scanning unit; 210: First 3D laser scanner; 220: Non-contact odometer; 230: Track gauge; 300: Handheld 3D laser scanning unit; 400: Processing unit. DETAILED DESCRIPTION

[0061] The following is a detailed description with reference to the accompanying drawings.

[0062] In the field of three-dimensional modeling of urban rail infrastructure, the application of laser scanning technology faces multiple technical challenges.

[0063] First, inefficient data acquisition is a core issue. Traditional scanning equipment often requires a significant amount of time to complete data acquisition when processing large-scale scenarios such as long tunnels. The complex structure and confined space of tunnels significantly reduce scanning efficiency. Furthermore, the high-precision scanning strategies employed to improve resolution further compress acquisition speed, creating an irreconcilable conflict between accuracy and efficiency. This technical bottleneck directly restricts the urban rail system's ability to respond to the demand for efficient modeling.

[0064] Second, the size and weight of the equipment limit its potential for application in complex environments. Current mainstream scanning equipment is generally bulky and overweight, making it difficult to flexibly deploy in space-constrained scenarios such as urban rail tunnels. Equipment movement requires specialized transportation or manual handling, which not only increases operational difficulty but also may limit the scanning angle due to spatial constraints, thus affecting the integrity of data collection. These physical limitations severely undermine the adaptability and operational efficiency of laser scanning technology in urban rail operation and maintenance scenarios.

[0065] Third, insufficient data acquisition quality and stability pose a significant challenge to modeling accuracy. In complex environments such as tunnels and underground projects, laser scanners are susceptible to line-of-sight limitations and variations in surface reflectivity. For example, the high reflectivity of metal components in tunnels can cause signal overload, while the low reflectivity of concrete walls can lead to data loss. These interfering factors collectively result in noise points, voids, or geometric distortion in the acquisition results, forcing technicians to invest significant effort in data cleaning, completion, and error correction. This post-processing not only prolongs the modeling cycle but also severely tests the reliability of the model.

[0066] To address the shortcomings of the existing technology, the present invention provides a system and method for rapidly collecting tunnel point cloud data. The present invention also provides a processor for building a three-dimensional model of a tunnel. The present invention also provides a storage medium storing an encoding program for processing first point cloud data and second point cloud data. The present invention also provides a rail vehicle 100 equipped with an onboard three-dimensional laser scanning unit 200 for use in conjunction with a handheld three-dimensional laser scanning unit 300 to collect point cloud data within the tunnel and construct a three-dimensional model of the tunnel.

[0067] Example 1

[0068] like Figures 18 to 20 As shown in Figure 1, urban rail line tunnels are underground or elevated tunnel structures between two adjacent stations in the urban rail transit system, primarily used for train passage and connecting different stations. They are a vital component of the urban rail system, fulfilling multiple functions such as train operation, passenger evacuation, and equipment installation.

[0069] Limitations of Conventional Tunnel Scanning Methods: Traditional tunnel scanning methods rely on stand-mounted 3D laser scanners. While these devices can capture relatively complete geometric data of the scanned area, they require station-by-station deployment and measurement, resulting in low overall efficiency. In large or complex urban rail tunnels, single-point deployment and measurement methods struggle to meet the demands for rapid scanning and efficient data collection.

[0070] The present invention provides a tunnel point cloud data rapid acquisition system for executing the tunnel point cloud data rapid acquisition method of the present invention.

[0071] like Figure 1 As shown, the system includes a vehicle-mounted 3D laser scanning unit 200, a handheld 3D laser scanning unit 300, and a processing unit 400. The vehicle-mounted 3D laser scanning unit 200 and the processing unit 400 are connected to each other in a wired or wireless manner. The handheld 3D laser scanning unit 300 and the processing unit 400 are connected to each other in a wired and / or wireless manner.

[0072] Wired communication: The vehicle-mounted 3D laser scanning unit 200 and the handheld 3D laser scanning unit 300 are equipped with wired connection ports (such as USB ports, HDMI ports, etc.). When the vehicle-mounted 3D laser scanning unit 200 and the handheld 3D laser scanning unit 300 are close to the processing unit 400, data can be transmitted by plugging in signal cables.

[0073] Wireless communication: The vehicle-mounted 3D laser scanning unit 200 and the handheld 3D laser scanning unit 300 are equipped with wireless communication modules. When the vehicle-mounted 3D laser scanning unit 200 and the handheld 3D laser scanning unit 300 are connected to the processing unit 400 through the wireless communication modules, data can be transmitted via wireless signals (such as WiFi signals, Bluetooth signals, and infrared signals).

[0074] Preferably, the wired connection port and the wireless communication module on the vehicle-mounted 3D laser scanning unit 200 and the handheld 3D laser scanning unit 300 can be set at the same time, or can be set one at a time.

[0075] The vehicle-mounted three-dimensional laser scanning unit 200 scans point cloud data of the three-dimensional space inside the tunnel and records the scanned position, so that the point cloud data is associated with the position information to obtain first point cloud data.

[0076] When held by a worker, the handheld three-dimensional laser scanning unit 300 performs a supplementary scan on the blocked area and / or cross section in the tunnel to obtain second point cloud data.

[0077] The processing unit 400 constructs a three-dimensional model of the tunnel based on the first point cloud data and the second point cloud data. Preferably, the processing unit 400 is a CPU, a server, a dedicated integrated chip, or a remote data processing platform that is internally provided with a coding program of the point cloud data processing method of the present invention.

[0078] The present invention's rapid tunnel point cloud data acquisition system is primarily used for collecting point cloud data of equipment in urban rail transit tunnels. By integrating multiple vehicle-mounted and handheld 3D laser scanning units, the system efficiently captures high-quality 3D point cloud and image data, enabling rapid and comprehensive scanning of equipment in urban rail transit tunnels.

[0079] Rail operation vehicle 100 Figure 2 and Figure 21 As shown, Figure 21The middle car body is yellow. The track work vehicle 100 is a special vehicle designed for railway track maintenance, inspection and construction. It is usually equipped with a variety of sensors and measuring equipment for real-time collection of three-dimensional spatial data of the track and surrounding structures during movement. The track work vehicle 100 can move on a track (such as a subway track). The track work vehicle 100 includes a simple car body, a wheel group and a drive group. The drive group includes a variable frequency motor acceleration and deceleration device, which supports low-speed starting and high-speed operation. The wheel group includes double rims and no rims, which are used to compensate for track deviations. Preferably, the simple car body is T-shaped. The wheel group is arranged under the transverse rod of the simple car body. The drive group is arranged under or inside the transverse rod of the simple car body to drive the wheel group. Preferably, the drive group also includes a controller for allowing workers to control the drive group through a control terminal.

[0080] The rail vehicle 100 is provided with a detachable vehicle-mounted three-dimensional laser scanning unit 200. The detachable means may be a mechanical fixing means such as a bolt connection means or a clamping means. Figure 21 As shown, the vehicle-mounted 3D laser scanning unit 200 is installed on the top of the rail vehicle 100. This allows the vehicle-mounted 3D laser scanning unit 200 to have a wider field of view, thereby collecting richer point cloud data.

[0081] The vehicle-mounted 3D laser scanning unit 200 includes a first 3D laser scanner 210 and a non-contact odometer 220. The first 3D laser scanner 210 is mounted on the rail vehicle 100 and is used to scan point cloud data of equipment and structures within the three-dimensional space of the tunnel while the rail vehicle 100 is in motion. For example, the first 3D laser scanner 210 can be a Leica P40 3D laser scanner. The Leica P40 3D laser scanner has high-precision and high-density point cloud data acquisition capabilities, enabling it to quickly capture 3D spatial information within the tunnel and accurately measure tunnel equipment and structures.

[0082] The contactless odometer 220 is a non-contact odometer. Mounted on the rail vehicle 100, it records the vehicle's trajectory, synchronizing location information with point cloud data and improving scanning accuracy and consistency.

[0083] Preferably, the vehicle-mounted 3D laser scanning unit 200 also includes a track gauge 230 mounted on the rail working vehicle 100. The track gauge 230 is, for example, a laser rangefinder. The track gauge 230 monitors the distance between the tracks within the tunnel in real time, allowing the processing unit 400 to select first point cloud data that meets track safety and facility maintenance requirements based on the distance. Preferably, during the scanning process, the processing unit 400 selects first point cloud data that meets track safety and facility maintenance requirements based on the distance between the tracks.

[0084] The handheld three-dimensional laser scanning unit 300 is also the second three-dimensional laser scanner of the present invention. Figure 3 、 Figure 5 and Figure 22 As shown, when held by a worker, the handheld three-dimensional laser scanning unit 300 performs a supplementary scan on the blocked area and / or section in the tunnel to obtain second point cloud data.

[0085] Obstructed areas include equipment, walls, and other areas within the tunnel. Using the handheld 3D laser scanning unit 300 for supplementary scanning ensures full data collection of obstructed areas, entire sections, and equipment within the tunnel, filling in the data gaps caused by obstructions in traditional stand-up scanning.

[0086] Specifically, the data collection process of the present invention is as follows.

[0087] After completing the preliminary work preparation, the data collection method proposed in the present invention is based on the collaborative operation of the vehicle-mounted 3D laser scanning unit 200 and the handheld 3D laser scanning unit 300. Under the premise of ensuring that the existing operating facilities and construction safety are not affected, the tunnel structure and ancillary equipment of the urban rail transit line section are collected with high precision and full coverage of the 3D space point cloud data. Figure 4 and Figure 23 As shown in FIG, a worker is debugging the vehicle-mounted three-dimensional laser scanning unit 200 through the terminal. Figure 23 As shown in the figure, the terminal's display interface information includes point name, code, settings, menu, control point measurement, etc. Control point measurement also includes the number of measurement cycles and electrical measurement data.

[0088] In the specific operation process, the vehicle-mounted 3D laser scanning unit 200 is installed on the rail vehicle 100. The vehicle-mounted 3D laser scanning unit 200 includes a first 3D laser scanner 210 (Leica P40 3D laser scanner), a non-contact odometer 220, and a track gauge 230. It can also be equipped with core components such as an industrial-grade control module and data recording memory.

[0089] After the operation begins, the rail vehicle 100 runs at a constant speed at a preset speed. The first 3D laser scanner 210 continuously scans the tunnel cross-section at a high frequency, acquiring dense and accurate 3D point cloud data. The non-contact odometer 220 simultaneously records the trajectory of the rail vehicle 100 and associates this spatial position information with the point cloud data in real time, achieving precise alignment of the point cloud data with the actual spatial coordinates, generating the first point cloud data. Simultaneously, the processing unit 400 fuses the captured image data with the first point cloud data, providing key semantic support for subsequent modeling, annotation, and recognition.

[0090] In the actual application environment of tunnels, due to the complex tunnel structure, dense distribution of auxiliary equipment and many obstructions, it is difficult for a single vehicle-mounted three-dimensional laser scanning unit 200 to achieve full coverage data collection, especially for locations such as tunnel corners, the back of the equipment and narrow passages. Conventional scanning methods have problems such as limited viewing angle and data loss. To solve the above problems, the present invention simultaneously deploys a handheld three-dimensional laser scanning unit 300, and workers enter the obstructed area to carry out supplementary scanning. The handheld device has the characteristics of flexibility, easy operation and high precision, and can effectively cover areas that are difficult to reach with traditional scanning equipment. Through the data fusion algorithm of the vehicle and handheld, the supplementary scanning data can be automatically spliced ​​into the overall point cloud model, ultimately realizing full-section, blind-spot-free and high-precision data collection of the tunnel. During the entire scanning process, the handheld three-dimensional laser scanning unit 300 can perform real-time monitoring according to preset parameters, covering key indicators such as scanning range, sampling accuracy, operation trajectory and data coverage.

[0091] During the scanning process of the vehicle-mounted three-dimensional laser scanning unit 200, when data missing, abnormal point cloud density or blocked areas are detected, the processing unit 400 determines the location information of the supplementary scan and triggers an early warning prompt. At the same time, the location information of the supplementary scan is sent to the worker's terminal to guide the worker to use the handheld three-dimensional laser scanning unit 300 for supplementary scanning in time to ensure the integrity and consistency of the final collected data.

[0092] Preferably, the steps performed by the processing unit 400 include:

[0093] Divide the first point cloud data collected in real time into regular spatial units. Count the number of point cloud points within each spatial unit to determine whether the preset sampling density threshold is reached. Identify areas with severely insufficient point cloud coverage or no data and mark them as data-missing areas. Calculate local point cloud density to detect areas with low density (undersampling) or high density (noise, repeated scans). Combine historical data with empirical thresholds to identify abnormal areas.

[0094] For example, the processing unit 400 sets the first point cloud dataset as:

[0095] P={p i |p i ∈R 3 ,i=1,2,……,n}.

[0096] In the above formula, p i Represents a point of the first point cloud data, and i represents the sequence number or index of a single point in the first point cloud data set.

[0097] The processing unit 400 divides the space into a series of regular spatial units (voxels):

[0098] V={v j|j=1,2,……,m}.

[0099] In the above formula, each voxel v j Represents a cubic area in space; j represents the sequence number or index of the divided voxel in space.

[0100] The processing unit 400 counts the number of points in each voxel:

[0101] n j =|{p i ∈v j}|.

[0102] The processing unit 400 sets the preset sampling density threshold as n min , to determine whether the voxels are sufficiently sampled.

[0103] If n j <n min , v j Marked as data missing area.

[0104] The processing unit 400 calculates the point cloud density (number of points per unit volume) within the voxel:

[0105]

[0106] In the above formula, d j Indicates the point cloud density, VOL(v j ) represents the volume of the voxel.

[0107] Assume that the normal density range is [d min ,d max ], the processing unit 400 determines the abnormality based on historical data and experience thresholds.

[0108] If d j <d min or d j >d max , then v j It is an area of ​​abnormal density.

[0109] Processing unit 400 integrates these detection results to comprehensively assess data quality. If missing data, abnormal point cloud density, or obstructed areas are detected, processing unit 400 determines the need for rescanning. Processing unit 400 locates the center coordinates and spatial extent of the abnormal area using the tunnel coordinate system or the vehicle-mounted scanning trajectory coordinate system. Processing unit 400 also annotates specific information such as the tunnel section number and start and end mileage.

[0110] Specifically, the processing unit 400 predefines the data quality function Q(v j ) The comprehensive missing and density anomaly indicators are:

[0111]

[0112] The processing unit 400 determines the abnormal region set:

[0113] A={v j |Q(v j )≠0}.

[0114] For each abnormal voxel v j ∈A, the processing unit 400 calculates the center coordinates of the unscanned area:

[0115] c j =(x j ,y j ,z j )=center point(v j ).

[0116] Processing unit 400 prioritizes anomalies based on their severity (e.g., missing area size, density deviation, and occlusion impact range). It then packages the anomaly type, severity, coordinates of the re-scan area center, and recommended re-scan parameters into a warning message package. The format of the warning message package may include map markers, graphical representations, and text descriptions. Processing unit 400 transmits the warning information to a terminal used by on-site workers via a wireless communication module. The terminal can be a tablet or dedicated terminal device.

[0117] After data collection is complete, workers can use portable terminals to conduct a preliminary preview and quality verification of the point cloud data, quickly assessing whether it meets accuracy and coverage requirements. For areas with errors or data gaps, a second scan can be immediately initiated, effectively reducing the risk of rework due to data quality issues.

[0118] On this basis, the vehicle-mounted 3D laser scanning unit 200 and the handheld 3D laser scanning unit 300 are also respectively provided with a memory. The memory will store all collected 3D point cloud data, image data and related location information according to a unified classification standard. Preferably, the memory completes labeling and archiving based on metadata such as tunnel number, acquisition date and mileage section. When the vehicle-mounted 3D laser scanning unit 200 and the handheld 3D laser scanning unit 300 respectively establish communication with the processing unit 400 in a wired and / or wireless manner, the memory uploads all data to the processing unit 400 to provide high-quality, structured basic data support for subsequent data processing, modeling analysis, facility management and disease identification. Setting up a memory can avoid the loss of real-time data transmission when the communication signal in the tunnel is poor.

[0119] The data processing process of the processing unit 400 on the received data is as follows.

[0120] S100: The processing unit 400 removes noise points from the first point cloud data or the second point cloud data.

[0121] In actual operations, the first and second point cloud data acquired by vehicle-mounted and handheld 3D laser scanners usually contain noise data (such as surveying personnel, pipelines, evacuation channels, etc.). The presence of this noise data will significantly affect the quality of the point cloud data. In view of the characteristics of tunnel point cloud data, a spatial circle fitting method based on the random sampling consensus algorithm (RANSAC) is used to remove noise data. The RANSAC algorithm calculates the parameters of the sphere using a small amount of point cloud data (the first point cloud data or the second point cloud data) and verifies the sphere parameters using the remaining point cloud data. After a sufficient number of iterative operations, the accurate parameters of the sphere are finally obtained.

[0122] S110: The processing unit 400 calculates and verifies the sphere parameters based on the random sampling consistency algorithm and the first point cloud data or the second point cloud data.

[0123] Let the equation of the sphere be:

[0124] (x-a0) 2 +(y-b0) 2 +(z-c0) 2 =R 2 .

[0125] In the above formula, (a0, b0, c0) represents the coordinates of the center of the sphere, (x, y, z) represents the coordinates of the segment point cloud data; R represents the radius of the sphere.

[0126] S120: The processing unit 400 calculates the center coordinates and radius of the segment point cloud data based on the conditional equation and the indirect adjustment method.

[0127] Let the equation of the plane passing through the center of the sphere be:

[0128] aa0+bb0+cc0=1.

[0129] In the above equation, (a0, b0, c0) represent the coordinates of the sphere's center, and (a, b, c) represent the parameters of the corresponding plane equation. The constant 1 represents the constraints imposed by the plane on the coordinates of the sphere's center, known as the conditional equation.

[0130] Based on the conditional equation and the indirect adjustment method, the processing unit 400 calculates the center coordinates and radius of the segment point cloud data, such as Figure 9 As shown. Figure 9 In the figure, the white arc lines on the blue background represent the calculated circles representing the pipe segment slices.

[0131] Define the observation vector:

[0132] L = (x1, y1, z1,..., x n , y n , z n ) T .

[0133] In the above formula, n represents the number of observation points in the observation value vector, i.e., the number of three-dimensional points in the first point cloud data used for calculation or fitting.

[0134] Assuming that there is an error in the measurement value, the unknown parameters need to be solved by indirect adjustment (error equation adjustment):

[0135] X = (a0, b0, c0, R) T .

[0136] Observation: f i (X) = (x i -a0) 2 +(y i -b0) 2 +(z i -c0) 2 , i = 1,..., m.

[0137] In the above formula, m represents the number of observation equations, i.e., the number of observation functions participating in indirect adjustment.

[0138] The condition equation is: g(X) = aa0+b-b0+cc0-1 = 0.

[0139] Introducing the parameter correction amount ΔX and the observation value correction amount W, the observation equation is linearized as:

[0140] f(X+ΔX)≈f(X)+AΔX=0.

[0141]

[0142] The linearized condition equation is: g(X+ΔX)≈g(X)+BΔX=0.

[0143]

[0144] The processing unit 400 constructs a least squares problem with a condition equation:

[0145] The conditions are: W+AΔX=-f(X);BΔX+g(X)=0.

[0146] The Lagrange multiplier method is used to introduce the multiplier λ, and the normal equation set is constructed:

[0147]

[0148] In the above formula, T represents the matrix transpose.

[0149] The processing unit 400 iteratively solves and calculates X (k+1) =X k +ΔX (k) , until the correction amount is small enough or converges.

[0150] The spherical parameters of the pipe segment slice finally estimated by the processing unit 400 are:

[0151] S130: The processing unit 400 calculates the deviation distance between the center coordinates of the segment point cloud and the center coordinates of the sphere.

[0152] Assume that the center coordinates of the segment point cloud are The coordinates of the center of the sphere are (a0, b0, c0), and the deviation distance d between the two is expressed as:

[0153]

[0154] S140: The processing unit 400 compares the deviation distance with the radius of the sphere, and eliminates points whose difference is greater than a difference threshold to eliminate noise points.

[0155] If |dR|<τ, the corresponding points are eliminated.

[0156] In the above formula, τ is the set difference threshold.

[0157] S200: The processing unit 400 performs data splicing on the first point cloud data and the second point cloud data at different positions to obtain spliced ​​standard point cloud data.

[0158] The stitching of the first and second point cloud data is a process of integrating the first and second point cloud data acquired at different scanning positions into a complete 3D model through spatial alignment and data fusion. This process includes key steps such as point cloud registration, parameter optimization, and error correction to ensure the final model has high accuracy and spatial consistency.

[0159] Before starting to stitch, Figure 10 As shown, the present invention imports the first point cloud data and the second point cloud data from different scanning positions into the processing unit 400 and performs a preliminary check to ensure that the data is complete and has no obvious noise.

[0160] S210: The processing unit 400 performs data calibration on the first point cloud data and the second point cloud data based on a known calibration target, such as Figures 13 to 15 shown. Figures 13 to 15 A diagram showing a portion of the point cloud data processing process according to the present invention.

[0161] The first step is data calibration, which ensures that the first point cloud data P (1)and the second point cloud data P (2) It can be interfaced with the actual coordinate system. The processing unit 400 can choose to use known calibration targets (such as spherical calibration targets, calibration boards, etc.) for geometric calibration, ensuring the correct position of the data in space.

[0162] Specifically,

[0163] In the above formula, P (1) represents the first point cloud data, P (2) represents the second point cloud data, represents the calibrated first point cloud data, represents the calibrated second point cloud data, T calib represents the calibration transformation matrix calculated based on the calibration target, C(·) represents the calibration function that applies the calibration transformation.

[0164] S220: The processing unit 400 performs coordinate system conversion on the data-calibrated first and second point cloud data to obtain third and fourth point cloud data in the global coordinate system.

[0165] After data calibration is complete, the first and second point cloud data are usually still in the local coordinate system and need to be converted to the global coordinate system.

[0166] The processing unit 400 can convert the first and second point cloud data from the local coordinate system of the scanner to the third and fourth point cloud data in the global coordinate system (such as WGS84 or UTM coordinate system) to be compatible with other data sets or geographic information systems (GIS).

[0167] Specifically,

[0168] In the above formula, P (3) represents the third point cloud data, P (4) represents the fourth point cloud data, T local→global represents the transformation matrix from the local coordinate system to the global coordinate system; μ(·) represents the coordinate transformation function.

[0169] S230: The processing unit 400 performs coordinate alignment based on position information to achieve registration of the third and fourth point cloud data at different positions; minimizes the geometric distance between the two point clouds based on the iterative closest point algorithm.

[0170] After coordinate system conversion is complete, the point cloud registration phase is entered. Registration is the core step of the splicing step of the present application, and the purpose is to align the point cloud data from different positions.

[0171] The processing unit 400 supports two main registration methods: automatic registration and manual registration.

[0172] Automatic registration typically relies on the processing unit 400 identifying and matching feature points (such as corners, edges, and planes) in the point cloud data. Using an algorithm (the Iterative Closest Point (ICP) algorithm), the processing unit 400 calculates the relative positions of the individual scans, automatically aligning the point cloud data at different locations to a common reference frame. During the automatic registration process, the processing unit 400 performs error analysis to ensure a high enough match between the data.

[0173] The core goal of the Iterative Closest Point (ICP) algorithm is to minimize the geometric distance between two point clouds. This is typically achieved by computing the optimal rigid transformation (including rotation and translation) between the source and target point clouds. The basic idea behind each iteration of the ICP algorithm is to gradually reduce the error between the point clouds by finding the "nearest neighbor" points between the source and target point clouds and then computing the optimal transformation matrix based on these paired points.

[0174] Assumption: The source point cloud point set is:

[0175] The target point cloud point set is:

[0176] S231: Find corresponding point pairs between the source point cloud and the target point cloud.

[0177] For each point cloud p in the source point cloud i , ICP will find the target point cloud q that is closest to it in the target point cloud i The distance metric is generally Euclidean distance. i The calculation formula is:

[0178]

[0179] In the above, ‖·‖2 represents the Euclidean distance.

[0180] S232: Calculate the optimal rigid transformation (rotation and translation).

[0181] For each pair of matching points (p i ,q i ), the goal of the ICP algorithm is to find a rigid transformation R and t so that the source point cloud best matches the target point cloud through this transformation. The transformation consists of a rotation matrix D and a translation vector t.

[0182] The goal is to minimize the square error between the transformed source point cloud and the target point cloud:

[0183]

[0184] In the above formula, D represents the rotation matrix (orthogonal matrix), t represents the translation vector, and E(D,t) represents the point cloud matching error function.

[0185] S233: Minimize the error.

[0186] By minimizing the error function E(D,t) above, we can get the optimal rotation matrix D and translation vector t. This process can be completed in the following two steps:

[0187] S2331: Calculate the rotation matrix D.

[0188] The rotation matrix D can be calculated by solving a least squares problem while minimizing the error function. The most common algorithm uses SVD (Singular Value Decomposition) to extract the rotation matrix from the covariance matrix of the point pair.

[0189] Calculate the centroid of the source point cloud as:

[0190] The centroid of the target point cloud is:

[0191] Construct a centralized point set:

[0192] Compute the covariance matrix:

[0193] Perform singular value decomposition (SVD) on the covariance matrix H: H = U∑V T .

[0194] In the above formula, U∑V T is the SVD decomposition matrix.

[0195] The rotation matrix is ​​calculated as: D = VU T .

[0196] S2332: Calculate the translation vector.

[0197] The translation vector t is calculated by the center of mass of the source point cloud and the target point cloud using the following formula:

[0198]

[0199] S234: Apply transformation.

[0200] Apply the calculated rotation matrix D and translation vector t to the source point cloud, and the updated source point cloud is:

[0201] S235: Iterative update.

[0202] By iterating steps S231 to S234 , the error between the source point cloud and the target point cloud is gradually reduced until the error converges or a predetermined maximum number of iterations is reached.

[0203] The goal of the iterative closest point (ICP) algorithm is to gradually reduce the geometric error between the point clouds by finding the nearest neighbor points between the source point cloud and the target point cloud, calculating the rotation matrix D and translation vector t, and then applying these transformations to the source point cloud.

[0204] The core of each iteration is to minimize the following error function:

[0205]

[0206] The stopping condition is: |E (k) -E (k-1) |<ε; or k=k max .

[0207] In the above formula, k represents the number of iterations, ε represents the error threshold, and k max Indicates the maximum number of iterations.

[0208] This process adjusts the source point cloud through multiple iterations until the best registration with the target point cloud is achieved. Figure 11 and Figure 12 The standard point cloud data obtained after stitching is shown.

[0209] S240: The processing unit 400 adjusts the standard point cloud data based on a global optimization algorithm.

[0210] If the automatic registration results are biased or inaccurate for certain areas, the user can perform manual registration. Manual registration typically involves fine-tuning the data by selecting specific registration points (such as corners or pipe joints) within the point cloud data, manually adjusting the data position to ensure precise alignment across all scan locations.

[0211] The registered standard point cloud data is usually not completely accurate and requires global optimization. The processing unit 400 performs final adjustments to the point cloud data using a global optimization algorithm (such as bundle adjustment). This algorithm comprehensively considers the standard point cloud data of all registration points, minimizes the overall error, and optimizes the spatial relationship of the data to ensure that the stitched standard point cloud data is as accurate and consistent as possible.

[0212] The calculation formula for optimizing the overall error is:

[0213] In the above formula, Z represents the global optimization parameter, f i,j (Z) represents the registration error function between the i-th and j-th standard point clouds.

[0214] S250: The processing unit 400 corrects the stitching points with large errors in the standard point cloud data.

[0215] After completing the global optimization, error analysis and correction are performed.

[0216] By using the error evaluation tool of the processing unit 400 , the errors between different scanning positions can be checked, and unreasonable stitching points can be corrected or the registration strategy can be adjusted.

[0217] Finally, the stitched standard point cloud data enters the data optimization and cleaning phase. Processing unit 400 provides functions such as noise filtering and point cloud thinning to remove unnecessary noise points, reduce the amount of standard point cloud data, and improve the efficiency of subsequent processing. The cleaned standard point cloud data can be compressed as needed or exported to standard CAD or BIM formats (such as DXF, DWG, .e57, etc.) for subsequent modeling, analysis, or visualization applications.

[0218] Throughout the point cloud stitching process, the processing unit 400 ensures high-precision stitching through a combination of automation and manual work, combined with advanced registration algorithms and error correction tools. Ultimately, the stitched standard point cloud data provides complete and accurate 3D spatial information.

[0219] This method utilizes the Iterative Closest Point (ICP) algorithm to efficiently register and stitch the collected first and second point cloud data. By calculating the optimal transformation matrix, the error between the source and target point clouds is gradually reduced, ensuring high accuracy and consistency in the stitched standard point cloud data. This method can rapidly process large amounts of point cloud data, providing efficient and accurate 3D reconstruction results.

[0220] S300: The processing unit 400 constructs a three-dimensional model of the tunnel based on the standard point cloud data.

[0221] like Figure 16 and Figure 17 As shown in FIG, the processed standard point cloud data is used for the real-time monitoring platform of subway rail transit data to monitor the tunnel.

[0222] like Figures 6 to 8 As shown, in the three-dimensional models constructed at different angles, the structure of the tunnel segments is clear, there are no interference points, and the image quality is clear.

[0223] The present invention achieves fast and efficient scanning of subway tunnels through the collaborative operation of the vehicle-mounted 3D laser scanning unit 200 and the handheld 3D laser scanning unit 300. Compared with traditional mounted 3D laser scanners, the mobile 3D laser scanning system has significant advantages:

[0224] First, measurement efficiency is significantly improved. The vehicle-mounted 3D laser scanning unit 200 can cover a larger area in a shorter time, breaking away from the limitations of traditional station-by-station measurement and significantly shortening the overall operation cycle.

[0225] Secondly, it has a stronger ability to adapt to complex obstructed environments. For blind spots caused by obstructions such as equipment and walls within tunnels, handheld scanning technology can be combined to effectively supplement data collection, ensuring the integrity and accuracy of full-section tunnel data.

[0226] Finally, data collection is more efficient. Leveraging the high-speed mobility of the railcar, the onboard 3D laser scanning unit 200 can quickly complete large-scale scans, reducing manual intervention and operational complexity, significantly improving operational efficiency and data reliability.

[0227] It should be noted that the above-mentioned specific embodiments are exemplary, and those skilled in the art can come up with various solutions inspired by the disclosure of the present invention, and these solutions also fall within the scope of the disclosure of the present invention and fall within the scope of protection of the present invention. Those skilled in the art should understand that the present invention specification and its drawings are illustrative and do not constitute a limitation on the claims. The scope of protection of the present invention is defined by the claims and their equivalents. The present invention specification contains multiple inventive concepts, such as "preferably" and "according to a preferred embodiment", which means that the corresponding paragraph discloses an independent concept, and the applicant reserves the right to file a divisional application based on each inventive concept.

Claims

1. A tunnel point cloud data rapid acquisition system, characterized in that: The system comprises: A vehicle-mounted three-dimensional laser scanning unit (200) scans point cloud data of the three-dimensional space inside the tunnel and records the position of the train, so that the point cloud data is associated with the position information to obtain first point cloud data; A handheld three-dimensional laser scanning unit (300) is used by a worker to perform a supplementary scan of the blocked area and / or cross section in the tunnel to obtain second point cloud data; A processing unit (400) constructs a three-dimensional model of the tunnel based on the first point cloud data and the second point cloud data.

2. The system according to claim 1, wherein: The vehicle-mounted three-dimensional laser scanning unit (200) comprises: A first three-dimensional laser scanner (210) is mounted on the rail vehicle (100) and is used to scan point cloud data of equipment and structures in the three-dimensional space inside the tunnel while the rail vehicle (100) is traveling; A non-contact odometer (220) is mounted on a rail operation vehicle (100) and is used to record the running track of the rail operation vehicle (100), thereby synchronizing position information and point cloud data.

3. The system according to claim 1 or 2, characterized in that The vehicle-mounted three-dimensional laser scanning unit (200) further includes a track gauge (230) mounted on the rail working vehicle (100). The track gauge (230) monitors the distance of the tracks in the tunnel in real time, so that the processing unit (400) can filter the first point cloud data that meets the requirements of track safety and facility maintenance based on the distance.

4. The system according to any one of claims 1 to 3, characterized in that: The processing unit (400) is configured to: Eliminating noise points from the first point cloud data or the second point cloud data; Splicing the first point cloud data and the second point cloud data at different positions to obtain spliced ​​standard point cloud data; A three-dimensional model of the tunnel is constructed based on the standard point cloud data.

5. The system according to any one of claims 1 to 4, characterized in that: The step of removing noise points from the first point cloud data or the second point cloud data by the processing unit (400) comprises: Calculate and verify the parameters of the sphere based on the random sampling consistency algorithm and the first point cloud data or the second point cloud data; Calculate the center coordinates and radius of the segment point cloud based on the conditional equation and indirect adjustment method; Calculate the deviation distance between the center coordinates of the segment point cloud and the center coordinates of the sphere; The deviation distance is compared with the radius of the sphere, and points with a difference greater than a difference threshold are eliminated to eliminate noise points.

6. The system according to any one of claims 1 to 5, characterized in that: The step of the processing unit (400) performing data splicing on the first point cloud data and the second point cloud data at different positions comprises: Performing data calibration on the first point cloud data and the second point cloud data based on a known calibration target; Performing coordinate system conversion on the first point cloud data and the second point cloud data after data calibration to obtain third point cloud data and fourth point cloud data in the global coordinate system; Aligning the coordinates of the third point cloud data and the fourth point cloud data at different positions based on the position information to achieve registration; Minimize the geometric distance between two point clouds based on the iterative closest point algorithm.

7. The system according to any one of claims 1 to 6, characterized in that: The step of the processing unit (400) performing data splicing on the first point cloud data and the second point cloud data at different positions further includes: Adjust standard point cloud data based on global optimization algorithm; Correct the stitching points with large errors in the standard point cloud data.

8. A method for rapid acquisition of tunnel point cloud data, characterized in that: The method comprises: Using a vehicle-mounted three-dimensional laser scanning unit (200) to scan point cloud data of the three-dimensional space inside the tunnel and record the position of the train, so that the point cloud data is associated with the position information to obtain first point cloud data; When held by a worker, a handheld three-dimensional laser scanning unit (300) is used to perform a supplementary scan on the blocked area and / or cross section in the tunnel to obtain second point cloud data; The processing unit (400) constructs a three-dimensional model of the tunnel based on the first point cloud data and the second point cloud data.

9. The method according to claim 8, characterized in that The method further comprises: Noise points are removed from the first point cloud data or the second point cloud data; the first point cloud data and the second point cloud data at different positions are spliced ​​to obtain spliced ​​standard point cloud data; and a three-dimensional model of the tunnel is constructed based on the standard point cloud data.

10. The method according to claim 8 or 9, characterized in that The method further comprises: Adjust standard point cloud data based on global optimization algorithm; Correct the stitching points with large errors in the standard point cloud data.

Images