A point cloud stitching method and device based on a high-precision coordinate measuring machine

Abstract

The application discloses a point cloud splicing method and device based on a high-precision coordinate measuring machine. The point cloud splicing method comprises the following steps: controlling a depth camera to pre-scan a measured object; adopting a path planning algorithm to plan scanning sites and scanning paths of a laser scanner, and determining coordinate position parameters of the scanning sites; controlling the laser scanner to collect multi-view point cloud data of the measured object according to the set scanning sites and scanning paths, and obtaining multi-view point cloud data; calculating a transformation matrix corresponding to each scanning site, and performing position transformation on each point cloud data; and splicing the multi-view point cloud data after position transformation and removing repeated points. The application is simple in steps and avoids problems such as incomplete point cloud data and large multi-site point cloud splicing errors caused by occlusion and inaccurate external parameters of the laser scanner, and greatly improves the accuracy and efficiency in the point cloud splicing process.

Description

Technical Field

[0001] This invention relates to point cloud stitching technology, and particularly to a point cloud stitching method and device based on a high-precision coordinate measuring machine. Background Technology

[0002] In today's era of rapid digital and information technology development, point cloud stitching technology, as a key component of 3D spatial data processing, has demonstrated its indispensable importance in multiple fields. Point clouds, as a precise 3D data representation of object surface morphology, are increasingly widely used in high-end technology fields such as autonomous driving, robot navigation, and geographic information systems. With the continuous advancement of deep learning technology, point cloud stitching technology has also achieved significant development, including aspects such as point cloud calibration, fusion, automatic classification, visualization, and post-processing.

[0003] In academic research, point cloud stitching technology has been recognized as a key technology in 3D imaging. This process involves finding the optimal spatial transformation matrix T to minimize the sum of distances between the transformed point cloud P and the target point cloud Q, achieving optimization across six degrees of freedom. Currently, the most commonly used point cloud stitching techniques rely on advanced algorithms, such as the ICP (Iterative Closest Point) algorithm, which achieves high-precision stitching by iteratively optimizing the correspondence between point clouds. Its advantages include the ability to handle point cloud data with partially overlapping regions and high stitching accuracy. Furthermore, feature-based point cloud stitching technology has also attracted considerable attention. It achieves stitching by extracting and matching salient feature points from the point cloud, offering strong robustness against noise and outliers.

[0004] In the current technological field, point cloud data stitching methods can be broadly categorized into manual registration and automatic registration, depending on whether human intervention is involved. In manual registration, the stitching process relies on identifying and establishing correspondences between corresponding points. This approach requires at least three sets of corresponding points to establish accurate registration transformations. However, this method demands high accuracy in point selection and is cumbersome, resulting in low overall registration efficiency. On the other hand, automatic registration uses algorithms to automatically extract feature points from point cloud data and performs registration and stitching between point clouds based on these feature points. Nevertheless, automatic registration still lags behind manual registration in terms of stability and accuracy, and its adaptability and flexibility in different application scenarios are also insufficient.

[0005] Current research on improving point cloud stitching technology focuses on using optimization algorithms to enhance the accuracy of point cloud stitching, thereby better obtaining the spatial transformation matrix required for point cloud stitching. However, when acquiring point clouds in complex scenes, the acquisition of extrinsic parameters often encounters inaccuracies, leading to significant errors in the stitching of point cloud data. In such cases, traditional algorithm-based stitching methods may not provide satisfactory results, often necessitating manual stitching to correct these errors. However, due to the subjectivity and complexity of manual operation, this method is not only inefficient but also prone to significant errors. Furthermore, errors from manual operation may also originate from multiple stages such as instrument calibration and data preprocessing, resulting in deviations between the stitched point cloud data and the geometry of the actual object. How to efficiently and accurately acquire the extrinsic parameters for point cloud stitching remains a challenge in current research. These challenges limit the application of point cloud stitching technology in high-precision detection and complex scene reconstruction. Summary of the Invention

[0006] Current point cloud stitching algorithms often fail to stitch due to poor initial image positioning, significant background occlusion, or complex scenes. These algorithms frequently get trapped in local maxima. To overcome these shortcomings, this invention provides a point cloud stitching method based on a high-precision coordinate measuring machine. This method eliminates the need for coarse and fine stitching steps, completing the acquisition and stitching of 3D point cloud data for a complete object in a single step. The method is simple and avoids the problems of incomplete point cloud data caused by occlusion and inaccurate extrinsic parameters in laser scanners, large errors in multi-site point cloud stitching, and difficulties in processing complex, large-scale point cloud data. This innovation in point cloud stitching technology for complex scenes significantly improves the accuracy and efficiency of the point cloud stitching process.

[0007] Compared to existing technologies, this invention represents a significant technological breakthrough. Existing stitching methods based on the ICP algorithm, while capable of stitching, rely on iterative optimization of point cloud correspondences. This results in massive computational demands when processing large-scale point cloud data, and stitching efficiency decreases sharply with increasing data volume. Furthermore, when the initial point cloud positions deviate significantly or occlusion exists, the method is prone to getting trapped in local optima, limiting stitching accuracy. While feature-based point cloud stitching techniques are robust to noise and outliers, the accuracy and stability of feature extraction are affected in complex scenarios, thus impacting stitching accuracy.

[0008] From a mathematical modeling perspective, this invention constructs a more accurate coordinate transformation model by precisely determining three translation parameters (X-axis translation parameter Xi, Y-axis translation parameter Yi, and Z-axis translation parameter Zi) and three rotation parameters (X-axis rotation parameter ωx, Y-axis rotation parameter ωy, and Z-axis rotation parameter ωz) of the scanning station relative to the reference point. This model is directly based on the precise position information of the scanning station. Compared to traditional stitching methods that rely on iterative calculations to find the optimal transformation matrix, it fundamentally avoids the problem of accumulated stitching errors caused by poor initial conditions. This provides a solid mathematical foundation for achieving high-precision point cloud stitching, simultaneously solving both the accuracy and efficiency issues of point cloud stitching.

[0009] The present invention also provides a point cloud stitching device, which uses a high-precision slide rail, a multi-degree-of-freedom robotic arm, a depth camera and a laser scanner to collect point cloud data. The technical means are simple and easy to implement.

[0010] The objective of this invention is achieved through the following technical solution:

[0011] This invention provides a point cloud stitching method based on a high-precision coordinate measuring machine, comprising the following steps:

[0012] S1 controls the depth camera to pre-scan the object under test, obtaining the position information of occluders in the scene and the outline of the object under test.

[0013] Based on the pre-scan results, S2 uses a path planning algorithm to plan the scanning stations and scanning paths of the laser scanner and determine the coordinate position parameters of each scanning station.

[0014] S3 sets the scanning station and scanning path according to the planning results of step S2, and controls the laser scanner to collect cloud data of the object under test from multiple perspectives according to the set scanning station and scanning path, and obtains point cloud data from multiple perspectives.

[0015] S4 obtains the transformation matrix of the point cloud data corresponding to each scanning station based on the coordinate position parameters of each scanning station obtained in step S2, and performs position transformation on each point cloud data obtained in step S3.

[0016] S5 stitches together the multi-view point cloud data after the position transformation in step S4 to obtain three-dimensional point cloud data, and removes duplicate points from the three-dimensional point cloud data.

[0017] Preferably, step S2 involves using a path planning algorithm to plan the scanning stations and scanning path of the laser scanner based on the pre-scanning results, and determining the coordinate position parameters of each scanning station. Specifically:

[0018] S21 processes the pre-scanning results obtained in step S1, converting the contour information and occlusion information of the measured object obtained in the pre-scanning into point cloud data.

[0019] S22 uses a greedy coverage algorithm. It selects the point that can cover the most unscanned contours as the first station from the starting point, marks the covered area, and continues to select the next station that can cover the most unscanned contours until the entire contour of the object being measured is covered.

[0020] S23 performs path planning, setting stations as graph nodes, connections as edges and weighting them, finding the shortest path sequence and checking its feasibility in conjunction with robot motion constraints, and designing a dynamic adjustment mechanism. If an anomaly is encountered, the RRT algorithm is used to replan the local path around the obstacle to the target station.

[0021] S24 uses all the finally determined stations as scanning stations and records their coordinate position parameters as extrinsic parameters for point cloud stitching.

[0022] Preferably, the step 24 of finding the shortest path sequence specifically involves using the A* or Dijkstra algorithm to find the shortest path sequence.

[0023] Preferably, for the i-th scanning station, its coordinate position parameters include the X-axis translation parameter Xi, the Y-axis translation parameter Yi, the Z-axis translation parameter Zi, and the X-axis rotation parameter ω relative to the reference point. x Y-axis rotation parameter ω relative to the reference point y Z-axis rotation parameter ω relative to the reference point z Where i = 1, 2, 3... represents the number of scanning stations.

[0024] Preferably, in step S4, the translation matrix, rotation matrix and transformation matrix of the i-th scanning station are as shown in equations (1), (2) and (3);

[0025]

[0026] Where i = 1, 2, 3, ..., j = x, y, z.

[0027] Preferably, the process of removing duplicate points from the 3D point cloud data in step S5 specifically involves:

[0028] By dividing the data into grid cells, detecting and removing coordinate points smaller than a set threshold to eliminate duplicate points, the data is then optimized and smoothed.

[0029] The present invention also provides a point cloud stitching device for implementing the point cloud stitching method based on a high-precision coordinate measuring machine, comprising a depth camera, a laser scanner, a multi-degree-of-freedom robotic arm, a high-precision slide rail, and a control device; the depth camera and laser scanner are mounted on the multi-degree-of-freedom robotic arm; the multi-degree-of-freedom robotic arm is mounted on the slide rail;

[0030] The depth camera is used to pre-scan the object being measured and store the scan data;

[0031] The laser scanner is used to collect, store, and process multi-view point cloud data of the object being measured.

[0032] The control device is used to plan high-precision slide rail paths, set the motion trajectory of multi-degree-of-freedom robotic arms, and send commands to multi-degree-of-freedom robotic arms to control their movement.

[0033] Preferably, the positioning accuracy of the high-precision slide rail is at the micrometer level.

[0034] The present invention also provides an electronic device, including a processor and a memory;

[0035] The memory is used to store non-transitory computer instructions;

[0036] When the non-temporal computer instructions are executed by the processor, the processor implements the point cloud stitching method based on a high-precision coordinate measuring machine.

[0037] The present invention also provides a storage medium for storing non-transitory computer instructions, which, when executed, perform the point cloud stitching method based on a high-precision coordinate measuring machine.

[0038] In this invention, the X-axis translation parameter, Y-axis translation parameter, and Z-axis translation parameter respectively characterize the translation distance difference of the point cloud data in the X, Y, and Z axes relative to the reference point; the X-axis rotation parameter, Y-axis rotation parameter, and Z-axis rotation parameter respectively characterize the rotation angle difference of the scanner position in the X, Y, and Z axes relative to the reference point. The X, Y, and Z axes are perpendicular to each other, the Z-axis is vertical, and the Y-axis is oriented towards the zero-degree direction of the laser scanner.

[0039] In this invention, the multi-degree-of-freedom robotic arm is connected to a high-precision slide rail. Following a pre-set path, it automatically adjusts and flexibly moves around the object being measured using a laser scanner, enabling omnidirectional scanning of the object. The multi-degree-of-freedom robotic arm possesses high flexibility, high precision, and strong adaptive capabilities. Its multiple degrees of freedom allow it to quickly and accurately complete various complex tasks in complex and ever-changing working environments. Furthermore, the multi-degree-of-freedom robotic arm endows the depth camera and laser scanner mounted on its head with high mobility, enabling precise movement and data acquisition at any position within the workspace.

[0040] In this invention, the depth camera has high resolution and low latency imaging performance, and can stably collect information under different lighting conditions, pre-scan the object under test, efficiently obtain the rough position of the scan point, and reduce the data collection time.

[0041] In this invention, the high-precision slide rail is a high-precision, high-rigidity, and low-friction slide rail with excellent linear motion accuracy and stability. It can maintain micron-level positioning accuracy during high-speed movement and can accurately acquire position information. The high-precision slide rail not only ensures the smooth movement of the scanning device, but also provides extremely high positioning accuracy, which can accurately acquire position information, enabling the laser scanner to maintain consistency and accuracy when collecting point cloud information.

[0042] In this invention, the laser scanner has high-speed scanning capability, high resolution and excellent measurement accuracy, and can acquire accurate point cloud data in a short time.

[0043] Compared with the prior art, the present invention has the following advantages and beneficial effects:

[0044] (1) The point cloud stitching method based on a high-precision coordinate measuring machine of the present invention firstly pre-scans the object under test using a depth camera to obtain the position information of occluders in the scene and the outline of the object under test; based on the pre-scanning results, a path planning algorithm is used to plan the scanning stations and scanning paths of the laser scanner to determine the coordinate position parameters of each scanning station; the laser scanner collects cloud data of the object under test from multiple perspectives according to the set scanning stations and scanning paths to obtain point cloud data from multiple perspectives; since the present invention accurately determines the three translation and three rotation amounts of the scanning stations based on the reference point during the path planning process, the transformation matrix corresponding to each scanning station is obtained accordingly; after the laser scanner obtains the point cloud data of each scanning station, the position transformation can be directly performed according to the transformation matrix, and a complete three-dimensional point cloud data can be obtained in one stitch. The method of the present invention does not require the steps of coarse stitching and fine stitching, the steps are simple, and the accuracy and efficiency are high.

[0045] (2) The point cloud stitching method based on a high-precision coordinate measuring machine of the present invention pre-scans the object under test using a depth camera to obtain the position information of occluders in the scene and the outline of the object under test. Then, based on the pre-scan results, a path planning algorithm is used to plan the scanning stations and scanning path of the laser scanner, determine the coordinate position parameters of each scanning station, and comprehensively consider factors such as scanning efficiency, coverage, and collision avoidance to optimize the scanning path. Through precise coordinate transformation and correction, the accurate position of the scanning points in three-dimensional space is obtained, providing the necessary prerequisite for high-precision scanning.

[0046] (3) The point cloud stitching device based on a high-precision coordinate measuring machine of the present invention adopts a depth camera, a high-precision slide rail, a multi-degree-of-freedom robotic arm, and a laser scanner. It effectively solves the problems of incomplete point cloud data caused by occlusion and inaccurate external parameters of the stand-up laser scanner, large point cloud stitching error at multiple stations, difficulty in processing complex and large-scale point cloud data, and difficulty in meeting the high-precision data requirements of the detection process. The technical means are simple and easy to implement. Attached Figure Description

[0047] Figure 1 This is a schematic diagram of the composition of a point cloud stitching device based on a high-precision coordinate measuring machine, according to an embodiment of the present invention.

[0048] Figure 2 This is a flowchart of a point cloud stitching method based on a high-precision coordinate measuring machine, according to an embodiment of the present invention. Detailed Implementation

[0049] The present invention will be further described in detail below with reference to the embodiments, but the implementation of the present invention is not limited thereto.

[0050] Some embodiments of the present invention provide a point cloud stitching device. For example... Figure 1 As shown, the point cloud stitching device includes a depth camera 4, a laser scanner 3, a multi-degree-of-freedom robotic arm 2, a high-precision slide rail 1, and a control device. The depth camera 4 and the laser scanner 3 are mounted on the multi-degree-of-freedom robotic arm 2. The multi-degree-of-freedom robotic arm 2 is mounted on the high-precision slide rail 1. The depth camera is used to pre-scan the object under test and store the scan data. The laser scanner is used to collect, store, and process multi-view point cloud data of the object under test. The control device (which may be a computer) is used to plan the high-precision slide rail path, set the motion trajectory of the multi-degree-of-freedom robotic arm, and send commands to the multi-degree-of-freedom robotic arm to control its movement.

[0051] In this embodiment, the high-precision slide rail is not only a high-precision, high-rigidity, and low-friction slide rail, but also has excellent linear motion accuracy and stability, and can maintain micron-level positioning accuracy during high-speed movement; the multi-degree-of-freedom robotic arm can automatically adjust and flexibly move around the object being measured to the laser scanner, giving the depth camera and laser scanner mounted on the robotic arm head a high degree of mobility, thereby realizing precise movement and data acquisition at any position in the workspace.

[0052] Other embodiments of the present invention provide a point cloud stitching method based on a high-precision coordinate measuring machine. For example... Figure 2 As shown, the point cloud stitching method includes the following steps:

[0053] S1 controls the depth camera to pre-scan the object under test, obtaining the position information of occluders in the scene and the outline of the object under test.

[0054] Based on the pre-scan results, S2 uses a path planning algorithm to plan the scanning stations and scanning paths of the laser scanner and determine the coordinate position parameters of each scanning station.

[0055] S3 sets the scanning station and scanning path according to the planning results of step S2, and controls the laser scanner to collect cloud data of the object under test from multiple perspectives according to the set scanning station and scanning path, and obtains point cloud data from multiple perspectives.

[0056] S4 calculates the transformation matrix corresponding to each scanning station based on the coordinate position parameters of each scanning station obtained in step S2, and performs position transformation on each point cloud data obtained in step S3.

[0057] S5 stitches together the multi-view point cloud data after the position transformation in step S4 to obtain three-dimensional point cloud data, and removes duplicate points from the three-dimensional point cloud data.

[0058] The following is a detailed description of the point cloud stitching method based on a high-precision coordinate measuring machine according to a specific embodiment of the present invention:

[0059] The point cloud stitching method based on a high-precision coordinate measuring machine includes the following steps:

[0060] S1 controls the depth camera to pre-scan the object under test, obtaining the position of occluders in the complex scene and the outline of the object under test, so as to carry out subsequent path planning.

[0061] By using a depth camera to perform a preliminary comprehensive scan of the object under test, a preliminary outline of the object is established, aiming to roughly obtain information on the possible locations of scan points; this stage provides basic data for subsequent precise scanning and path planning.

[0062] Based on the pre-scan results, S2 uses a path planning algorithm to plan the scanning stations and scanning path of the laser scanner, and determines the coordinate position parameters of each scanning station; in this embodiment, the specific implementation is as follows:

[0063] S21 processes the pre-scanning results obtained in step S1, converting the contour information and occlusion information of the measured object obtained in the pre-scanning into point cloud or polygon form.

[0064] S22 employs a greedy coverage algorithm, adjusting the laser scanner to a suitable distance from the object being measured. It selects the point that covers the most unscanned contours as the first station, marks the covered area, and continues to select the next station that covers the most unscanned contours until the entire contour of the object being measured is covered. For occlusions, it prioritizes stations with minimal impact. When there is a large area of ​​occlusion, it first determines the surrounding auxiliary stations to obtain edge information and then plans the internal stations.

[0065] S23 performs path planning, setting stations as graph nodes, connections as edges and weighting them, finding the shortest path sequence and checking its feasibility in conjunction with robot motion constraints, and designing a dynamic adjustment mechanism. If an anomaly is encountered, the RRT algorithm is used to replan the local path around the obstacle to the target station.

[0066] S24 uses all the finally determined stations as scanning stations and records their coordinate position parameters as extrinsic parameters for point cloud stitching.

[0067] For the i-th (i = 1, 2, 3, ...) scanning station, its coordinate position parameters include the X-axis translation parameter Xi, the Y-axis translation parameter Yi, the Z-axis translation parameter Zi, and the X-axis rotation parameter ω relative to the reference point. x Y-axis rotation parameter ω relative to the reference point y Z-axis rotation parameter ω relative to the reference point z ;

[0068] The translation matrix, rotation matrix, and transformation matrix of the i-th scanning station are shown in equations (1), (2), and (3);

[0069]

[0070]

[0071] Where i = 1, 2, 3, ..., j = x, y, z.

[0072] S3 sets the scanning station and scanning path according to the planning results of step S2, and controls the laser scanner to collect cloud data of the object under test from multiple perspectives according to the set scanning station and scanning path, and obtains point cloud data from multiple perspectives.

[0073] S4 calculates the transformation matrix corresponding to each scanning station based on the coordinate position parameters of each scanning station obtained in step S2, and performs position transformation on each point cloud data obtained in step S3.

[0074] S5 stitches together the multi-view point cloud data after the position transformation in step S4 to obtain three-dimensional point cloud data, and removes duplicate points from the three-dimensional point cloud data; in this embodiment, the specific method for removing duplicate points from the three-dimensional point cloud data is as follows:

[0075] By dividing the data into grid cells, detecting and removing coordinate points smaller than a set threshold to eliminate duplicate points, the data is then optimized and smoothed.

[0076] Embodiments of the present invention also provide an electronic device, including a processor and a memory;

[0077] The memory is used to store non-transitory computer instructions;

[0078] When the non-temporary computer instructions are executed by the processor, the processor implements the point cloud stitching method based on a high-precision coordinate measuring machine.

[0079] Embodiments of the present invention also provide a storage medium for storing non-transitory computer instructions, which, when executed, perform the point cloud stitching method based on a high-precision coordinate measuring machine.

[0080] The above embodiments are preferred embodiments of the present invention, but the embodiments of the present invention are not limited to the embodiments described above. Any changes, modifications, substitutions, combinations, or simplifications made without departing from the spirit and principle of the present invention shall be considered equivalent substitutions and shall be included within the protection scope of the present invention.

Claims

1. A point cloud stitching method based on a high-precision coordinate measuring machine, characterized in that, The method comprises the following steps: S1, controlling a depth camera to pre-scan a measured object to obtain position information of an occlusion in a scene and an outline of the measured object; S2, according to the pre-scan result, using a path planning algorithm to plan a scanning station and a scanning path of a laser scanner, and determining coordinate position parameters of each scanning station, specifically: S21, performing data processing on the pre-scan result obtained in step S1, and converting the outline information and the occlusion information of the measured object obtained by pre-scanning into point cloud data; S22, using a greedy covering algorithm to select a point that can cover the most unscanned outline as the first station from the starting point, mark the covered area, and continue to select the next station that can cover the most unscanned outline, until the entire outline of the measured object is covered; S23, performing path planning, setting the stations as graph nodes, connecting them as edges and weighting, finding the shortest path sequence and checking the feasibility in combination with the robot motion restriction, and designing a dynamic adjustment mechanism, such as using the RRT algorithm to re-plan the local path to bypass the obstacle to the target station when an exception is encountered; S24, taking all the finally determined stations as scanning stations, and recording their coordinate position parameters as external parameters for point cloud stitching; S3, setting the scanning station and the scanning path according to the planning result of step S2, and controlling the laser scanner to collect multi-view point cloud data of the measured object according to the set scanning station and scanning path, and obtaining multi-view point cloud data; S4, calculating a transformation matrix corresponding to each scanning station according to the coordinate position parameters of each scanning station obtained in step S2, and performing position transformation on each point cloud data obtained in step S3; S5, stitching the multi-view point cloud data after position transformation in step S4 to obtain three-dimensional point cloud data, and removing repeated points from the three-dimensional point cloud data.

2. The high-precision coordinate measuring machine based point cloud stitching method of claim 1, wherein, In step 23, the shortest path sequence is found by using A* or Dijkstra algorithm.

3. The high-precision coordinate measuring machine based point cloud stitching method of claim 1, wherein, For the i-th scanning station, its coordinate position parameters include X-axis translation parameter Xi of the scanning station relative to the reference point, Y-axis translation parameter Yi of the scanning station relative to the reference point, Z-axis translation parameter Zi of the scanning station relative to the reference point, X-axis rotation parameter of the scanning station relative to the reference point , Y-axis rotation parameter of the scanning station relative to the reference point, and Z-axis rotation parameter of the scanning station relative to the reference point; wherein i = 1, 2, 3…, represents the number of scanning stations.

4. The high-precision coordinate measuring machine based point cloud stitching method of claim 3, wherein, In step S4, the translation matrix, the rotation matrix and the transformation matrix of the i-th scanning station are shown in formulas (1), (2) and (3); （1） （2） （3） Wherein, i=1, 2, 3…, j=x, y, z.

5. The high-precision coordinate measuring machine based point cloud stitching method of claim 4, wherein, In step S5, the repeated points are removed from the three-dimensional point cloud data by dividing the data into grid units, detecting and removing coordinate points less than a set threshold to eliminate repeated points, and then performing data optimization and smoothing processing. The system comprises a depth camera, a laser scanner, a multi-degree-of-freedom robot arm, a high-precision slide rail, and a control device; the depth camera and the laser scanner are installed on the multi-degree-of-freedom robot arm; the multi-degree-of-freedom robot arm is installed on the slide rail; 6. A point cloud splicing device for implementing the point cloud splicing method based on a high-precision coordinate measuring machine according to any one of claims 1 to 5, characterized in that, The depth camera is used for pre-scanning a measured object and storing scanning data; The laser scanner is used for collecting, storing and processing multi-view point cloud data of the measured object; The control device is used for planning a high-precision slide rail path, setting a multi-degree-of-freedom robot arm motion trajectory, and sending instructions to the multi-degree-of-freedom robot arm to control its motion. The positioning accuracy of the high-precision slide rail is microns.

7. The point cloud stitching apparatus of claim 6, wherein, The system comprises a processor and a memory; 8. An electronic device, comprising: The memory is used for storing non-transitory computer instructions; ​ When the non-transitory computer instructions are executed by the processor, the processor implements the point cloud splicing method based on the high-precision coordinate measuring machine according to any one of claims 1-5.

9. A storage medium, characterized by A storage device for storing non-transitory computer instructions, when the non-transitory computer instructions are executed, performing the point cloud splicing method based on the high-precision coordinate measuring machine according to any one of claims 1-5.

Images