Multi-camera 3D Point Cloud Stitching Method and Measurement System Based on Dynamic Correction

Abstract

The present application discloses a multi-camera three-dimensional point cloud stitching method and a measurement system based on dynamic correction, belonging to the technical field of three-dimensional measurement, including: moving the point cloud reconstruction component along the planned solution path to solve the matrix conversion relationship between the feature point coordinate system and the scanning coordinate system; the point cloud reconstruction component measures the measurement target along the planned scanning path, and the multi-camera tracking component tracks the point cloud reconstruction component and the measurement platform to solve the matrix conversion relationship between the tracking coordinate system and the global coordinate system and the matrix conversion relationship between the scanning coordinate system and the tracking coordinate system, and converts the three-dimensional point cloud data of the measurement target measured by the point cloud reconstruction component into the same coordinate system; converting the three-dimensional point cloud of the measurement target corresponding to the current scanning path point scanned by the point cloud reconstruction component and the feature points of the feature polyhedron into the point coordinate data corresponding to the first scanning path point, and converting them into the global coordinate system to complete point cloud stitching. The present invention corrects the problem of accuracy loss in measurement.

Description

Technical Field

[0001] The present application relates to a multi-camera three-dimensional point cloud stitching method and a measurement system based on dynamic correction, belonging to the technical field of three-dimensional measurement. Background Art

[0002] Point cloud stitching is a key step in three-dimensional reconstruction and measurement, which is a process of stitching and fusing point clouds obtained from multiple different perspectives or positions into a complete three-dimensional model. According to different stitching principles, point cloud stitching is divided into multiple methods such as reference point stitching, feature stitching, deep learning matching stitching, etc. For reference point stitching, the characteristics are high matching accuracy but high dependence on the environment, and reference points need to be arranged on the target surface; for feature stitching, the characteristics are relatively high matching, but it requires obvious features on the measurement target surface; for deep learning matching stitching, the characteristics are low dependence on the environment and features, but the stitching effects of different iterative matching algorithms vary greatly. Among them, reference point stitching is currently recognized in the industry as a point cloud stitching method with relatively high accuracy. Since the reference points are always on or near the measurement target, they will affect the actual shape of the measurement target, so the applicable range is limited.

[0003] To get rid of the influence of reference points and achieve high-precision point cloud stitching, there have emerged various point cloud stitching methods in the industry that remove the dependence on reference points. Among them, to meet the needs of industrial production and detection, corresponding automated detection technical solutions have been generated by combining point cloud stitching methods that remove the dependence on reference points. The currently mature technical means include: there is a rigid connection between the reference points and the measurement target, and the reference points do not need to be attached to the measurement target. The point cloud reconstruction component captures the measurement target and the reference points near the target at multiple positions, and the binocular tracking component tracks the feature points on the shell of the point cloud reconstruction component at the corresponding positions. Currently, the commonly used point cloud reconstruction component is a three-dimensional laser scanner. Based on the principle of laser ranging, by projecting a single beam or multiple laser lines onto the surface of the measurement target, the three-dimensional point cloud data of the measured target can be quickly reconstructed. Using the matrix transformation relationship between the scanning coordinate system of the point cloud reconstruction component and the feature point coordinate system of the feature points on the shell of the point cloud reconstruction component, the matrix transformation relationship between the feature point coordinate system of the feature points on the shell of the point cloud reconstruction component and the tracking coordinate system of the binocular tracking component, and the matrix transformation relationship between the tracking coordinate system of the binocular tracking component and the global coordinate system where the measurement plane with the measurement target is located. The point cloud data of the measurement target and the three-dimensional coordinates of the reference points obtained by the point cloud reconstruction component are transformed from the scanning coordinate system of the point cloud reconstruction component to the global coordinate system of the measurement plane, and the point cloud stitching of the measurement target is completed. Common automatic tracking systems are as Figure 1 shown.

[0004] However, there may be the following three problems:

[0005] 1) Precision loss problem caused by environmental vibration: The vibration in the industrial site affects the relative position relationship between the binocular tracking component and the measurement plane, and this position relationship directly affects the stitching precision of the measured target point cloud captured. For the offset problem of this position relationship, methods such as foundation backfilling and vibration isolation platform earthquake resistance have been given in the industry, but the cost is relatively high.

[0006] 2) Precision loss problem caused by temperature change: The environmental temperature in the workshop where the measurement system is located may change continuously, and at the same time, the system's own operation will also cause self-heating. Among them, the structural components fixing the cameras will deform under the action of heat. This deformation will directly change the relative position relationship between the cameras of the binocular tracking component, and this position relationship directly determines the reconstruction precision of the three-dimensional point cloud data of the measurement target. Although for the above problems, methods such as selecting structural components with small thermal expansion and contraction to fix the cameras or performing temperature compensation in real time by algorithms have been given in the industry, the overall cost is relatively high. Among them, the method of temperature compensation is complex to operate and may have the situation that the temperature compensation amount is not accurate enough.

[0007] 3) Precision loss problem caused by limited tracking perspective: The observation points of the binocular tracking component are fixed, and its field of view is limited, so the effective measurement range is limited. For the perspective problem, methods such as performing station transfer shooting on the binocular tracking component have also been given in the industry to solve it, but such a tracking method will cause cumulative errors between different observation points. Summary of the Invention

[0008] According to one aspect of the present application, a multi-camera three-dimensional point cloud stitching method based on dynamic correction is provided, and this method dynamically corrects the precision loss problems caused by environmental vibration and temperature change during measurement.

[0009] The multi-camera three-dimensional point cloud stitching method based on dynamic correction is characterized by including:

[0010] Step 1: Calibrate the multi-eye tracking component, the point cloud reconstruction component, the feature points on the point cloud reconstruction component, and the measurement platform where the measurement target is located, and establish a tracking coordinate system, a scanning coordinate system, a feature point coordinate system, and a global coordinate system;

[0011] Step 2: Move the point cloud reconstruction component along the planned solution path to solve the matrix transformation relationship between the feature point coordinate system and the scanning coordinate system;

[0012] Step 3: The point cloud reconstruction component measures the measurement target along the planned scanning path, and tracks the feature points on the point cloud reconstruction component and the backlight marking points on the measurement platform through the multi-eye tracking component, and solves the matrix transformation relationship between the tracking coordinate system and the global coordinate system and the matrix transformation relationship between the scanning coordinate system and the tracking coordinate system under each scanning path;

[0013] Step 4: While photographing the measurement target and the feature points on the feature polyhedron through the point cloud reconstruction component, convert the three-dimensional point cloud data of the measurement target obtained by the point cloud reconstruction component into coordinate values in the same reference coordinate system; wherein, both the measurement target and the feature polyhedron are located on the electric turntable of the measurement platform.

[0014] Step 5: Convert the three-dimensional point cloud of the measurement target and the coordinate data of the feature points of the feature polyhedron in the scanning coordinate system corresponding to the current scanning path points scanned by the point cloud reconstruction component to the point coordinate data in the scanning coordinate system corresponding to the first scanning path point, and complete point cloud stitching in the global coordinate system.

[0015] Further, the said Step 2 includes:

[0016] Step 2.1: The point cloud reconstruction component runs to the th solution path point, photographs the calibration board and determines the transformation relationship between the scanning coordinate system and the calibration board coordinate system.

[0017] Step 2.2: The multi-view tracking coordinate system determines the transformation relationship between the tracking coordinate system and the feature point coordinate system by photographing the feature points on the point cloud reconstruction component.

[0018] Step 2.3: Calculate the transformation relationship from the scanning coordinate system to the feature point coordinate system through the transformation relationship from the calibration board coordinate system to the scanning coordinate system and the transformation relationship from the feature point coordinate system to the tracking coordinate system.

[0019] Further, Step 2.1 includes:

[0020] Suppose the three-dimensional point coordinates of the common marker points on the calibration board in the scanning coordinate system photographed at the th solution path point are denoted as , then:

[0021] ;

[0022] Wherein, and respectively represent the rotation matrix and translation matrix from the calibration board coordinate system to the scanning coordinate system at the th solution path point, is the three-dimensional coordinate of the common marker point corresponding to in the calibration board coordinate system; is the transformation matrix from the calibration board coordinate system to the scanning coordinate system at the current scanning path point.

[0023] Further, Step 2.2 includes:

[0024] Suppose at the For the solution path points, the three-dimensional coordinates of the feature points on the point cloud reconstruction component in the tracking coordinate system obtained by the multi-view tracking component are denoted as , then:

[0025] ;

[0026] Among them, represents the number of solution path points, i ∈ {1, 2, 3, …, n}, and are respectively the rotation matrix and translation matrix from the feature point coordinate system to the tracking coordinate system under the -th solution path point; is the three-dimensional coordinate of the common feature point corresponding to in the feature point coordinate system; is the transformation matrix from the feature point coordinate system to the tracking coordinate system under the current solution path point; is the transformation matrix from the feature point coordinate system to the tracking coordinate system.

[0027] Furthermore, step 2.3 includes:

[0028] Each solution path point satisfies:

[0029] ;

[0030] For all solution path points, the matrix conversion relationship remains unchanged. Therefore, all path points satisfy:

[0031] ;

[0032] Simplified representation:

[0033] ;

[0034] Among them:

[0035] ;

[0036] ;

[0037] ;

[0038] Determine the matrix conversion relationship from the calibration board coordinate system to the tracking coordinate system under all solution path points ; Determine the matrix conversion relationship from the scanning coordinate system to the feature point coordinate system under all scanning path points .

[0039] Further, in step 3, dynamically solve the matrix transformation relationships between the tracking coordinate system and the global coordinate system, and between the scanning coordinate system and the tracking coordinate system for each scanning path, including:

[0040] Step 3.1: The multi-camera tracking component tracks the backlight marker points on the measurement platform at each scanning path point, and dynamically corrects the transformation relationship from the tracking coordinate system to the global coordinate system at the current path point;

[0041] Suppose the three-dimensional coordinates of the common marker points on the backlight marker points in the tracking coordinate system obtained at the current scanning path point are , then:

[0042] ;

[0043] Among them, and are the rotation matrix and translation matrix for transforming the three-dimensional coordinates of the backlight marker points captured by the multi-camera tracking component at the th scanning path point in the global coordinate system to the tracking coordinate system; represents the matrix transformation relationship from the global coordinate system to the tracking coordinate system at the current scanning path point; is the three-dimensional coordinates of the corresponding common marker points in the global coordinate system;

[0044] Step 3.2: The multi-camera tracking component simultaneously tracks the feature points on the point cloud reconstruction component at the same scanning path point, and dynamically solves the transformation relationship from the feature point coordinate system on the point cloud reconstruction component to the tracking coordinate system at the current scanning path point;

[0045] Suppose the three-dimensional coordinates of the common feature points of the feature points in the tracking coordinate system obtained by the multi-camera tracking component at the current scanning path point are , then:

[0046] ;

[0047] Among them, and are the rotation matrix and translation matrix for transforming the three-dimensional coordinates of the feature points captured by the multi-camera tracking component at the th path point in the feature point coordinate system to the tracking coordinate system; represents the matrix transformation relationship from the feature point coordinate system to the tracking coordinate system at the current path point; is the three-dimensional coordinates of the feature points on the point cloud reconstruction component captured by the multi-camera tracking component at the th path point in the feature point coordinate system.

[0048] Further, in step 4, the point cloud reconstruction component simultaneously photographs the measurement target and the feature points on the feature polyhedron, and unifies the three-dimensional point cloud data of the measurement target obtained by the point cloud reconstruction component into one coordinate system, including:

[0049] Step 4.1: The point cloud reconstruction component photographs the point cloud data of the measurement target and the feature polyhedron with feature points in the scanning coordinate system corresponding to the first scanning path point; the point cloud reconstruction component continues to move along the scanning path and photographs the point cloud data of the measurement target and the feature polyhedron with feature points in the scanning coordinate system corresponding to the current scanning path point, where there are at least 4 common feature points in the point cloud data scanned by adjacent path points. ; the point cloud reconstruction component continues to move along the scanning path and photographs the point cloud data of the measurement target and the feature polyhedron with feature points in the scanning coordinate system corresponding to the current scanning path point , where there are at least 4 common feature points in the point cloud data scanned by adjacent path points;

[0050] Based on the common feature point coordinate data of the measurement target scanned by the point cloud reconstruction component at adjacent path points, solve the matrix transformation relationship between the two scanning coordinate systems, that is:

[0051] ;

[0052] where, represents the number of scanning path points, k ∈ {1, 2, 3, …, m}, is the coordinate data of the feature points on the feature polyhedron in the scanning coordinate system corresponding to the current path point scanned by the point cloud reconstruction component at the th path point; is the coordinate data of the feature points on the corresponding feature polyhedron in the scanning coordinate system corresponding to the current path point scanned by the point cloud reconstruction component at the th scanning path point; and are the rotation matrix and translation matrix for transforming the three-dimensional coordinates of the feature points on the feature polyhedron in the scanning coordinate system corresponding to the th scanning path point to the scanning coordinate system corresponding to the

[0053] Step 4.2: At the th scanning path point, the matrix transformation relationship between the scanning coordinate system corresponding to the point cloud reconstruction component and the scanning coordinate system corresponding to the first scanning path point is:

[0054] ;

[0055] where, is the Coordinate data of the feature points on the feature polyhedron in the scanning coordinate system corresponding to the current path point scanned by the point cloud reconstruction component under the scanning path point; It is the coordinate data of the feature points on the feature polyhedron in the scanning coordinate system corresponding to the current path point scanned by the point cloud reconstruction component under the first scanning path point; Indicates the matrix transformation relationship between the scanning coordinate systems corresponding to the

[0056] Determine the matrix transformation relationship between the three-dimensional point cloud data of the measurement target captured by the point cloud reconstruction component and the feature point data on the feature polyhedron in the scanning coordinate system corresponding to the th scanning path point in the scanning coordinate system corresponding to the first scanning path point:

[0057] ;

[0058] Among them, is the coordinate data of the three-dimensional point cloud of the measurement target and the feature points of the feature polyhedron in the scanning coordinate system corresponding to the current scanning path point scanned by the point cloud reconstruction component under the th scanning path point; is the coordinate data of the three-dimensional point cloud of the measurement target and the feature points of the feature polyhedron in the scanning coordinate system corresponding to the current scanning path point scanned by the point cloud reconstruction component under the first scanning path point.

[0059] Furthermore, in the fifth step, the conversion to the global coordinate system and the completion of point cloud stitching include:

[0060] ;

[0061] Among them, is the point coordinate data of the three-dimensional point cloud data of the measurement target scanned by the point cloud reconstruction component and the feature points of the feature polyhedron in the global coordinate system under the th scanning path point; is the inverse matrix of the matrix transformation relationship from the global coordinate system to the tracking coordinate system under the th scanning path point; is the matrix transformation relationship from the feature point coordinate system to the tracking coordinate system under the th scanning path point; is the matrix transformation relationship from the scanning coordinate system to the feature point coordinate system during the measurement process; is the point coordinate data of the three-dimensional point cloud of the measurement target and the feature points of the feature polyhedron in the scanning coordinate system corresponding to the th scanning path point in the scanning coordinate system corresponding to the first scanning path point;

[0062] There are at least 4 common feature points in the point cloud data of adjacent path points, and the point cloud stitching in the global coordinate system is realized through the common feature points.

[0063] Furthermore, it also includes:

[0064] Step 6: Package the three-dimensional point cloud data of the measured object after stitching to obtain a complete mesh data model of the measured object, and perform detection and comparison with the CAD design model corresponding to the measured object.

[0065] According to another aspect of the present application, there is provided a multi-camera three-dimensional measurement system based on dynamic correction, which is characterized by including:

[0066] A measurement platform, on which a measured object and backlight marking points are arranged;

[0067] A point cloud reconstruction component, which is installed on the measurement platform through a robotic arm, and the point cloud reconstruction component measures the measured object;

[0068] A multi-camera tracking component, which tracks and measures the measured object on the measurement platform. Feature points are arranged on the point cloud reconstruction component and are within the field of view of the multi-camera tracking component;

[0069] Wherein, the measured object is installed on the measurement platform through an electric turntable and is located at the center of the electric turntable. A plurality of feature polyhedrons with feature points are also arranged on the tabletop of the electric turntable, and the plurality of feature polyhedrons are located around the measured object;

[0070] The backlight marking points are located on the periphery of the electric turntable.

[0071] The beneficial effects that the present application can produce include:

[0072] The multi-camera three-dimensional point cloud stitching method and measurement system based on dynamic correction provided by the present application move the point cloud reconstruction component along the planned solution path, solve the matrix transformation relationship between the feature point coordinate system and the scanning coordinate system, and the point cloud reconstruction component measures the measured object along the planned scanning path. The multi-camera tracking component simultaneously tracks the feature points on the point cloud reconstruction component and the backlight marking points on the measurement platform, dynamically solves the matrix transformation relationship between the scanning coordinate system and the tracking coordinate system under the same path point, and quickly converts the three-dimensional point cloud data of the measured object measured by the point cloud reconstruction component at each path point to the global coordinate system to realize point cloud stitching. It dynamically corrects the accuracy loss problems caused by environmental vibration and temperature changes during measurement; introduces an electric turntable, reduces the spatial tracking range of the tracking component, and increases the system freedom to 7 degrees, meeting the high-precision and high-efficiency measurement requirements for large workpieces, and ensuring the point cloud stitching efficiency and accuracy. Description of the Drawings

[0073] Figure 1 It is a schematic diagram of an automatic tracking system in the prior art;

[0074] Figure 2 It is a flowchart of a multi-camera three-dimensional point cloud stitching method based on dynamic correction in an embodiment of the present application;

[0075] Figure 3 It is a schematic diagram of a multi-camera three-dimensional measurement system based on dynamic correction in an embodiment of the present application;

[0076] List of components and reference numerals: 1 - measurement platform; 2 - measurement target; 3 - backlight marking points; 4 - point cloud reconstruction component; 5 - robotic arm; 6 - multi-view tracking component; 7 - electric turntable; 8 - feature polyhedron. Detailed implementation manners

[0077] The present application will be described in detail below with reference to the embodiments, but the present application is not limited to these embodiments.

[0078] See Figure 2 , a multi-camera three-dimensional point cloud stitching method based on dynamic correction, characterized by including:

[0079] Step 1, calibrate the multi-view tracking component 6, the point cloud reconstruction component 4, the feature points on the point cloud reconstruction component 4, and the measurement platform where the measurement target is located, and establish a tracking coordinate system, a scanning coordinate system, a feature point coordinate system, and a global coordinate system;

[0080] Step 2, move the point cloud reconstruction component 4 along the planned solution path to solve the matrix transformation relationship between the feature point coordinate system and the scanning coordinate system;

[0081] Step 3, the point cloud reconstruction component 4 measures the measurement target 2 along the planned scanning path, and the multi-view tracking component 6 tracks the feature points on the point cloud reconstruction component 4 and the backlight marking points 3 on the measurement platform, and solves the matrix transformation relationship between the tracking coordinate system and the global coordinate system and the matrix transformation relationship between the scanning coordinate system and the tracking coordinate system for each scanning path;

[0082] Step 4, the point cloud reconstruction component 4 simultaneously takes pictures of the measurement target and the feature points on the feature polyhedron 8, and converts the three-dimensional point cloud data of the measurement target measured by the point cloud reconstruction component 4 into coordinate values in the same reference coordinate system; wherein, both the measurement target and the feature polyhedron 8 are located on the electric turntable 7 of the measurement platform;

[0083] Step 5: Convert the 3D point cloud of the measurement target and the coordinate data of the feature points of the feature polyhedron 8 corresponding to the current scanning path points obtained by the point cloud reconstruction component 4 in the scanning coordinate system corresponding to the first scanning path point to the global coordinate system and complete point cloud stitching.

[0084] Specifically, based on the principle of close-range photogrammetry, calibrate the multi-camera tracking component 6 to determine the initial internal and external parameters of the multi-camera and establish a tracking coordinate system; use the adjustment and optimization method to calibrate the point cloud reconstruction component 4 to determine the internal and external parameters of the camera of the point cloud reconstruction component 4 and establish a scanning coordinate system; solve the matrix conversion relationship between the scanning coordinate system and the feature point coordinate system, and the robotic arm 5 drives the point cloud reconstruction component 4 to move along the planned solution path. The point cloud reconstruction component 4 and the multi-camera tracking component 6 simultaneously photograph the ceramic calibration board. During the whole process, the tracking component does not move relative to the ceramic calibration board, and the feature points of the point cloud reconstruction component 4 and the shell of the point cloud reconstruction component 4 are rigidly connected to solve the matrix conversion relationship between the scanning coordinate system and the feature point coordinate system. The robotic arm 5 drives the point cloud reconstruction component 4 to perform 3D measurement along the planned scanning path. The electric turntable 7 fixed with the measurement target and the feature polyhedron 8 rotates or remains stationary according to the viewing angle requirements of the point cloud reconstruction component 4 at each path point. The multi-camera tracking component 6 fixed on the top of the system frame simultaneously tracks the feature points on the shell of the point cloud reconstruction component 4 and the backlight marking points 3 fixed on the measurement platform of the marble base. Dynamically correct the matrix conversion relationship between the tracking coordinate system and the global coordinate system. Affected by environmental vibration and system temperature change, the external parameters of the cameras of the multi-camera tracking component change, and this set of external parameter relationships directly affect the matrix conversion relationship between the tracking coordinate system and the global coordinate system. Through the 3D point coordinates of the marking points photographed at each scanning path point in the tracking coordinate system, combined with the 3D point coordinates of the corresponding marking points obtained by prior photogrammetry in the global coordinate system, dynamically correct the matrix conversion relationship between the tracking coordinate system and the global coordinate system at each scanning path point. Dynamically solve the matrix conversion relationship between the scanning coordinate system and the tracking coordinate system. The point cloud reconstruction component 4 moves along the planned scanning path driven by the robotic arm 5. The multi-camera tracking component 6 tracks the feature points on the shell of the point cloud reconstruction component 4 at each scanning path point to determine the 3D coordinates of the feature points photographed at the current path point in the tracking coordinate system, and combined with the matrix conversion relationship between the scanning coordinate system and the feature point coordinate system, solve the matrix conversion relationship between the scanning coordinate system and the tracking coordinate system at the current path point.

[0085] Further, driven by the robotic arm 5, the point cloud reconstruction component 4 moves along a predefined scanning path and captures the point cloud data of the measurement target on the electric turntable 7 and the feature polyhedron 8 with feature points at each path point. Considering that during the entire scanning path, the scanning coordinate system continuously changes with the path points of the current point cloud reconstruction component 4, and the electric turntable 7 also rotates according to different scanning path points, the relative position between the electric turntable 7 and the point cloud reconstruction component 4 changes accordingly. First, the point cloud data of the measurement target and the feature polyhedron 8 with feature points in the scanning coordinate system corresponding to the first scanning path point are captured; continue scanning, and capture the point cloud data of the measurement target and the feature polyhedron 8 with feature points in the scanning coordinate system corresponding to the current scanning path point. It is required that there are at least 4 common feature points on the known feature polyhedron 8 in the point cloud data of each path point in the current scanning coordinate system and the point cloud data in the corresponding scanning coordinate system of the adjacent path point; then, based on the coordinate data of the common feature points scanned from adjacent path points, determine the matrix transformation relationship between the scanning coordinate systems of all other scanning path points and the scanning coordinate system corresponding to the first scanning path point; finally, for the point cloud data captured at all path points in the current scanning coordinate system, apply the matrix transformation relationship between the scanning coordinate system of the corresponding scanning path point and the scanning coordinate system corresponding to the first scanning path point to uniformly transform the point cloud data scanned from all other scanning path points to the scanning coordinate system corresponding to the first scanning path point.

[0086] While solving the problem of the relative position change between the electric turntable 7 and the point cloud reconstruction component 4, based on the determined matrix transformation relationship between the tracking coordinate system and the global coordinate system at each path point, and the matrix transformation relationship between the scanning coordinate system and the tracking coordinate system, the point cloud data obtained by the point cloud reconstruction component 4 at each path point in the scanning coordinate system corresponding to the first scanning path point can be transformed to the global coordinate system and the point cloud stitching can be completed.

[0087] Preferably, a large field-of-view calibration algorithm based on close-range photogrammetry is used to calibrate the multi-camera tracking component 6, which specifically includes: the cameras of the multi-camera tracking component 6 simultaneously capture the honeycomb calibration board to obtain a set of calibration board images; move the position of the honeycomb calibration board on the measurement platform and repeat capturing the calibration board images in multiple poses. Given the global point data of the calibration board in its own coordinate system, combined with the captured calibration board images, first match the coded points in the images and preliminarily calculate and determine the initial values of the internal and external parameters of each camera; then match the non-coded points on the calibration board and perform bundle adjustment calculation to optimize and solve the initial values of the internal and external parameters of each camera.

[0088] The calibration of the point cloud reconstruction component 4 adopts the adjustment and optimization method, and the calibration plate is a ceramic calibration plate with circular coded points and non-coded points regularly arranged. Among them, the point cloud reconstruction component 4 is a grating structured light scanning component, which projects grating stripes with periodically changing thickness and phase on the surface of the object through the grating, and the camera interprets the deformed grating stripes modulated by the object surface, so as to determine and output the three-dimensional point cloud of the area where the grating is projected on the object surface. The point cloud reconstruction accuracy of this type of technical means is relatively high.

[0089] The second step includes:

[0090] Step 2.1: The point cloud reconstruction component 4 runs to the th solution path point, takes a picture of the calibration plate and determines the transformation relationship between the scanning coordinate system and the calibration plate coordinate system;

[0091] Step 2.2: The multi-view tracking coordinate system determines the transformation relationship between the tracking coordinate system and the feature point coordinate system by taking pictures of the feature points on the point cloud reconstruction component 4;

[0092] Step 2.3: Calculate the transformation relationship between the scanning coordinate system and the feature point coordinate system through the transformation relationship from the calibration plate coordinate system to the scanning coordinate system and the transformation relationship from the feature point coordinate system to the tracking coordinate system.

[0093] Step 2.1 includes:

[0094] Suppose the three-dimensional point coordinates of the common marker points on the calibration plate in the scanning coordinate system taken at the th solution path point are recorded as , then:

[0095] ;

[0096] Among them, and respectively represent the rotation matrix and translation matrix from the calibration plate coordinate system to the scanning coordinate system at the th solution path point, is the three-dimensional coordinate of the common marker point corresponding to in the calibration plate coordinate system; is the transformation matrix from the calibration plate coordinate system to the scanning coordinate system at the current scanning path point.

[0097] Step 2.2 includes:

[0098] Suppose at the th solution path point, the three-dimensional coordinates of the feature points on the point cloud reconstruction component 4 obtained by the multi-view tracking component 6 in the tracking coordinate system are recorded as , then:

[0099] ;

[0100] Among them, represents the number of solution path points, and \(i\in\{1,2,3,\cdots,n\}\), and are respectively the rotation matrix and translation matrix from the feature point coordinate system to the tracking coordinate system under the -th solution path point; is the three-dimensional coordinate of the common feature point corresponding to in the feature point coordinate system; is the transformation matrix from the feature point coordinate system to the tracking coordinate system under the current solution path point; solved by the SVD singular value decomposition method .

[0101] Step 2.3 includes:

[0102] Since the movement duration of the entire solution path does not exceed 5 minutes, the matrix conversion relationship between the two remains unchanged, and each solution path point satisfies:

[0103] ;

[0104] For all solution path points, the matrix conversion relationship remains unchanged. Therefore, each path point satisfies:

[0105] ;

[0106] Simplified representation:

[0107] ;

[0108] Among them:

[0109] ;

[0110] ;

[0111] ;

[0112] Determine the matrix conversion relationship from the calibration plate coordinate system to the tracking coordinate system under all solution path points ; Determine the matrix conversion relationship from the scanning coordinate system to the feature point coordinate system under all scanning path points .

[0113] In the third step described above, dynamically solve the matrix conversion relationship between the tracking coordinate system and the global coordinate system and the matrix conversion relationship between the scanning coordinate system and the tracking coordinate system for each scanning path, including:

[0114] Step 3.1: The multi-view tracking component 6 tracks the backlight marking points 3 on the measurement platform under each scanning path point, and dynamically corrects the conversion relationship between the tracking coordinate system and the global coordinate system under the current path point;

[0115] Suppose the three-dimensional coordinates of the common marking points on the backlight marking point 3 in the tracking coordinate system obtained under the current scanning path point are , then:

[0116] ;

[0117] Among them, and are the rotation matrix and translation matrix for transforming the three-dimensional coordinates of the backlight marking point 3 captured by the multi-view tracking component 6 at the th scanning path point in the global coordinate system to the tracking coordinate system; ;

[0118] represents the matrix conversion relationship from the global coordinate system to the tracking coordinate system under the current scanning path point; is the three-dimensional coordinates of the corresponding common marking points in the global coordinate system;

[0119] Step 3.2: The multi-view tracking component 6 simultaneously tracks the feature points on the point cloud reconstruction component 4 under the same scanning path point, and dynamically solves the conversion relationship between the feature points on the point cloud reconstruction component 4 and the tracking coordinate system in the feature point coordinate system under the current scanning path point;

[0120] Suppose the three-dimensional coordinates of the common feature points of the feature points in the tracking coordinate system obtained by the multi-view tracking component 6 under the current scanning path point are , then:

[0121] ;

[0122] Among them, and are the rotation matrix and translation matrix for transforming the three-dimensional coordinates of the feature points captured by the multi-view tracking component 6 at the th path point in the feature point coordinate system to the tracking coordinate system; ;

[0123] represents the matrix conversion relationship from the feature point coordinate system to the tracking coordinate system under the current path point; is the three-dimensional coordinates of the feature points on the shell of the point cloud reconstruction component 4 captured by the multi-view tracking component 6 at the th path point in the feature point coordinate system.

[0124] Specifically, the multi-view tracking component 6 is used to simultaneously track the feature points on the housing of the point cloud reconstruction component 4 and the backlight marking points 3 on the measurement platform, and dynamically solve the matrix transformation relationship between the tracking coordinate system and the global coordinate system and the matrix transformation relationship between the scanning coordinate system and the tracking coordinate system at each path point. The feature points on the outer housing of the point cloud reconstruction component 4 and the backlight marking points 3 on the measurement platform respectively obtain accurate three-dimensional global coordinates of the points through a preset high-resolution device, that is, the three-dimensional global coordinates of the feature points on the outer housing of the point cloud reconstruction component 4 in the feature point coordinate system are known; the three-dimensional global coordinates of the backlight marking points 3 on the measurement platform in the global coordinate system are known. Subsequent automated measurement processes do not require repeated measurements, unless the point cloud reconstruction component 4 or the marking points undergo large deformations resulting in a decrease in accuracy, in which case re-measurement is required. During the three-dimensional measurement process of the point cloud reconstruction component 4 along the planned scanning path, the feature polyhedron 8 and the measurement target are fixed on the electric turntable 7, and the electric turntable 7 will move into the scanning area of the point cloud reconstruction component 4 according to the scanning path points. In this application, the multi-view tracking component 6 simultaneously tracks the feature points on the housing of the point cloud reconstruction component 4 and the backlight marking points 3 on the measurement platform, thereby achieving high-precision tracking of the movement of the point cloud reconstruction component 4 and dynamic correction of the external parameters of the multi-view tracking component 6 at the corresponding scanning path points. During the tracking process, the multi-view tracking component 6 remains stationary. By tracking the feature points on the housing of the point cloud reconstruction component 4 and the marking points on the measurement platform, the matrix transformation relationship from the tracking coordinate system to the global coordinate system and the matrix transformation relationship from the scanning coordinate system to the tracking coordinate system are respectively calculated at the current path point.

[0125] Among them, a scanning coordinate system is established on the point cloud reconstruction component 4, a feature point coordinate system is established on the feature points of the housing of the point cloud reconstruction component 4, a calibration plate coordinate system is established on the ceramic calibration plate, a tracking coordinate system is established on the multi-view tracking component 6, and a global coordinate system is established on the marble measurement tabletop. A ceramic calibration plate with a small coefficient of thermal expansion is placed on the electric turntable 7 and remains stationary. The solution path of the robotic arm 5 is planned, and the robotic arm 5 drives the point cloud reconstruction component 4 to move. The point cloud reconstruction component 4 moves to the specified th movement path point and takes a picture of the ceramic calibration plate to determine the matrix transformation relationship between the scanning coordinate system and the calibration plate coordinate system ; the multi-view tracking component 6 takes pictures of the feature points on the point cloud reconstruction component 4 at the same path point to determine the matrix transformation relationship between the tracking coordinate system and the feature point coordinate system ; based on and these two sets of matrix transformation relationships, at the th path point, the matrix transformation relationship from the calibration plate coordinate system to the tracking coordinate system is established:

[0126] ;

[0127] Wherein: represents the matrix transformation relationship between the feature point coordinate system of the housing of the point cloud reconstruction component 4 and the tracking system coordinate system under the th path point. Given the global feature points, the multi - binocular tracking component 6 captures a set of feature point images at each path point, and solves the matrix transformation relationship between the feature point coordinate system and the tracking coordinate system for each path; represents the matrix transformation relationship between the scanning coordinate system and the feature point coordinate system, which is an unknown quantity; represents the matrix transformation relationship between the calibration plate coordinate system and the scanning coordinate system. Given the global calibration plate points, this matrix can be directly obtained.

[0128] For all path points, the matrix transformation relationship remains unchanged;

[0129] Solve the matrix transformation relationship from the calibration plate coordinate system to the tracking coordinate system for all path points ; At the same time, obtain the matrix transformation relationship from the scanning coordinate system to the feature point coordinate system that remains unchanged for all path points .

[0130] Based on the above - mentioned technical solution, preferably, the point cloud reconstruction component 4 performs three - dimensional measurement according to the planned scanning path. At each path point, the matrix transformation relationship between the dynamic correction tracking coordinate system and the global coordinate system, and the matrix transformation relationship between the scanning coordinate system and the tracking coordinate system are dynamically solved. The specific process is as follows: First, the multi - binocular tracking component 6 tracks the marked points on the measurement platform of the marble base under a single path point, and obtains the matrix transformation relationship between the tracking coordinate system and the global coordinate system under the current path point. At the same time, the multi - binocular tracking component 6 tracks the feature points on the housing of the point cloud reconstruction component 4 under the same path point, and obtains the matrix transformation relationship between the tracking coordinate system and the feature point coordinate system under the current path point.

[0131] In step 4 above, the point cloud reconstruction component 4 simultaneously captures the measurement target and the feature points on the feature polyhedron 8, and unifies the three - dimensional point cloud data of the measurement target obtained by the point cloud reconstruction component 4 into one coordinate system, including:

[0132] Step 4.1, the point cloud reconstruction component 4 captures the point cloud data of the measurement target and the feature polyhedron 8 with feature points in the scanning coordinate system corresponding to the first scanning path point ; The point cloud reconstruction component 4 continues to move along the scanning path and captures the point cloud data of the measurement target and the feature polyhedron 8 with feature points in the scanning coordinate system corresponding to the current scanning path point , wherein, there are at least 4 common feature points in the point cloud data scanned by adjacent path points;

[0133] Based on the common feature point coordinate data of the measurement target obtained by the component 4 for point cloud reconstruction during scanning at adjacent path points, solve the matrix transformation relationship between the two scanning coordinate systems, that is:

[0134] ;

[0135] Among them, represents the number of points of the scanning path, k ∈ {1, 2, 3, …, m}, is the coordinate data of the feature points on the feature polyhedron 8 in the scanning coordinate system corresponding to the current path point scanned by the component 4 for point cloud reconstruction at the -th scanning path point; is the coordinate data of the feature points on the corresponding feature polyhedron 8 in the scanning coordinate system corresponding to the current path point scanned by the component 4 for point cloud reconstruction at the -th scanning path point; and are the rotation matrix and translation matrix for transforming the three-dimensional coordinates of the feature points on the feature polyhedron 8 in the scanning coordinate system corresponding to the -th scanning path point to the scanning coordinate system corresponding to the -th scanning path point; ;

[0136] represents the matrix transformation relationship between the -th scanning coordinate system and the scanning coordinate system corresponding to the -th scanning path point at the current path point, represents a real number matrix, and 4 is the number of factors of the matrix;

[0137] Step 4.2. At the -th scanning path point, the matrix transformation relationship between the scanning coordinate system corresponding to the component 4 and the scanning coordinate system corresponding to the first scanning path point is:

[0138] ;

[0139] Among them, is the coordinate data of the feature points on the feature polyhedron 8 in the scanning coordinate system corresponding to the current path point scanned by the component 4 for point cloud reconstruction at the -th scanning path point; is the coordinate data of the feature points on the feature polyhedron 8 in the scanning coordinate system corresponding to the current path point scanned by the component 4 for point cloud reconstruction at the first scanning path point; represents the matrix transformation relationship between the scanning coordinate system corresponding to the -th scanning path point and the scanning coordinate system corresponding to the first path point;

[0140] Determine the Under the [[n]]-th scanning path point, the matrix transformation relationship between the three-dimensional point cloud data of the measurement target captured by the point cloud reconstruction component 4 and the feature point data on the feature polyhedron 8 in the scanning coordinate system corresponding to the first scanning path point:

[0141] ;

[0142] Among them, is the coordinate data of the three-dimensional point cloud of the measurement target and the feature points of the feature polyhedron 8 in the scanning coordinate system corresponding to the current scanning path point scanned by the point cloud reconstruction component 4 under the [[n]]-th scanning path point; is the coordinate data of the three-dimensional point cloud of the measurement target and the feature points of the feature polyhedron 8 in the scanning coordinate system corresponding to the current scanning path point scanned by the point cloud reconstruction component 4 under the first scanning path point.

[0143] In the fifth step, converting to the global coordinate system and completing point cloud stitching includes:

[0144] ;

[0145] Among them, is the point coordinate data of the three-dimensional point cloud data of the measurement target scanned by the point cloud reconstruction component 4 and the feature points of the feature polyhedron 8 in the global coordinate system under the [[n]]-th scanning path point; is the inverse matrix of the matrix transformation relationship from the global coordinate system to the tracking coordinate system under the [[n]]-th scanning path point; is the matrix transformation relationship from the feature point coordinate system to the tracking coordinate system under the [[n]]-th scanning path point; is the matrix transformation relationship from the scanning coordinate system to the feature point coordinate system during the measurement process; is the point coordinate data of the three-dimensional point cloud of the measurement target and the feature points of the feature polyhedron 8 in the scanning coordinate system corresponding to the [[n]]-th scanning path point in the scanning coordinate system corresponding to the first scanning path point;

[0146] There are at least 4 common feature points in the point cloud data of adjacent path points, and point cloud stitching in the global coordinate system is realized through the common feature points.

[0147] ​​​​​Specifically, the point cloud reconstruction component 4 simultaneously photographs the measurement target and the feature points on the feature polyhedron 8, and unifies the three-dimensional point cloud data of the measurement target measured by the point cloud reconstruction component 4 at each path point into one coordinate system. When the point cloud reconstruction component 4 performs three-dimensional measurement according to the planned path, the measurement target and the feature polyhedron 8 are fixed on the electric turntable 7 at the same time. The electric turntable 7 moves into the measurement range of the point cloud reconstruction component 4 according to the planned scanning path points, greatly reducing the spatial tracking range of the point cloud reconstruction component 4 and ensuring that the point cloud reconstruction component 4 can obtain high-precision point clouds.

[0148] Among them, driven by the robotic arm 5, the point cloud reconstruction component 4 moves along the established scanning path and photographs the point cloud data of the measurement target and the feature polyhedron 8 with feature points on the electric turntable 7 at each path point. Considering that in the entire scanning path, the scanning coordinate system changes continuously with the path points of the current point cloud reconstruction component 4, and the electric turntable 7 also rotates according to different scanning path points, the relative position between the electric turntable 7 and the point cloud reconstruction component 4 changes accordingly.

[0149] It is worth noting that the point cloud reconstruction component 4 photographs the point cloud data of the measurement target and the feature polyhedron 8 with feature points in the scanning coordinate system corresponding to the first scanning path point. ; Then, the point cloud reconstruction component 4 continues to move along the scanning path and photographs the point cloud data of the measurement target and the feature polyhedron 8 with feature points in the scanning coordinate system corresponding to the current scanning path point. It is required that there are at least 4 common feature points in the point cloud data scanned by adjacent path points; then, based on the common feature point coordinate data of the measurement target scanned by the point cloud reconstruction component 4 at adjacent path points, the matrix transformation relationship between the two scanning coordinate systems is solved.

[0150] Then, the coordinate data of the feature points on the feature polyhedron 8 in the scanning coordinate system corresponding to the current path point scanned by the point cloud reconstruction component 4 at each path point are unified into the scanning coordinate system corresponding to the first scanning path point, that is, the matrix transformation relationship between the scanning coordinate system corresponding to the point cloud reconstruction component 4 at the th path point and the scanning coordinate system corresponding to the first scanning path point is determined.

[0151] Finally, since the feature polyhedron 8 and the measurement target on the electric turntable 7 are in a good connection during the scanning process of the point cloud reconstruction component 4, the matrix transformation relationship between the scanning coordinate system corresponding to the th scanning path point and the scanning coordinate system corresponding to the first path point is applicable to the three-dimensional point cloud data of the measurement target and all the feature point data on the feature polyhedron 8. Applying the determined matrix transformation relationship, the Under a certain number of path points, the three-dimensional point cloud data of the measurement target obtained by the point cloud reconstruction component 4 and the coordinate data of the feature points on the feature polyhedron 8 in the scanning coordinate system corresponding to the first scanning path point.

[0152] Based on the above technical solution, preferably, the three-dimensional point cloud of the measurement target corresponding to the current path point scanned by the point cloud reconstruction component 4 and the point coordinate data of the feature point coordinates of the feature polyhedron 8 in the scanning coordinate system corresponding to the first scanning path point. Since there are at least 4 common feature points in the point cloud data of adjacent path points, point cloud stitching in the global coordinate system can be achieved through the common feature points.

[0153] It further includes:

[0154] Step six: Package the three-dimensional point cloud data of the stitched measurement target to obtain a complete mesh data model of the measurement target, and perform inspection and comparison with the CAD design model corresponding to the measurement target.

[0155] Specifically, triangulate and package the obtained three-dimensional point cloud data of the measurement target to obtain a complete mesh data model of the measurement target, and perform data post-processing, including filling holes and removing unnecessary edge regions. Inspect and compare the point cloud data of the measurement target and the imported CAD design model of the object to be measured. The inspections performed include calculating the surface deviation, section deviation, geometric tolerance, etc. of the actual object to be measured, and a detection report can be output.

[0156] See Figure 3 , a multi-camera three-dimensional measurement system based on dynamic correction, characterized by including:

[0157] A measurement platform, on which a measurement target and backlight marking points 3 are provided;

[0158] A point cloud reconstruction component 4, which is installed on the measurement platform through a robotic arm 5, and the point cloud reconstruction component measures the measurement target;

[0159] A multi-camera tracking component 6, which tracks and measures the measurement target on the measurement platform. Feature points are provided on the point cloud reconstruction component 4 and are within the field of view of the multi-camera tracking component 6;

[0160] Among them, the measurement target is installed on the measurement platform through an electric turntable 7 and is located at the center of the electric turntable 7. A plurality of feature polyhedrons 8 with feature points are also provided on the tabletop of the electric turntable 7, and the plurality of feature polyhedrons 8 are located around the measurement target;

[0161] The backlight marking points 3 are located on the peripheral side of the electric turntable 7.

[0162] Specifically, the point cloud reconstruction component 4 is fixed at the end of the robotic arm 5. Feature points are provided on the outer shell of the point cloud reconstruction component 4 and are within the field of view of the multi-camera tracking component 6. The multi-camera tracking component 6 is provided with a plurality of industrial cameras, and the cameras are connected by metal structural members. The electric turntable 7 is fixed within the working range of the robotic arm 5. A measurement target is fixed at the center of the electric turntable 7, and a feature polyhedron 8 with feature points is provided on the tabletop of the electric turntable 7. The ceramic calibration plate is placed on the electric turntable 7 for calibration and is within the field of view of the point cloud reconstruction component 4. The feature polyhedron 8 is fixed on the electric turntable 7 and is within the field of view of the point cloud reconstruction component 4. The backlight marking points 3 are fixed on the measurement platform of the marble base and are within the field of view of the multi-camera tracking component 6. The measurement platform of the marble base remains stationary, and backlight marking points 3 are provided on the surface.

[0163] The robotic arm 5 is used to drive the point cloud reconstruction component 4 to move along a planned path. The point cloud reconstruction component 4 is fixed at the end of the industrial robotic arm and is used to perform three-dimensional measurement. Feature points are fixed on the outer shell of the point cloud reconstruction component 4. The electric turntable 7 is fixed within the working range of the robotic arm 5 and is within the optimal measurement distance of the point cloud reconstruction component 4. The feature polyhedron 8 is provided on the tabletop of the electric turntable 7, and feature points are provided on the surface to facilitate the point cloud reconstruction component 4 to track the movement of the measurement target. The cameras of the multi-camera tracking component 6 are fixed on the metal structural members and are used to track the feature points on the outer shell of the point cloud reconstruction component 4 and the backlight marking points 3 on the measurement plane of the marble base. The measurement plane of the marble base is provided on the operation cabinet, and backlight marking points 3 are provided on the measurement plane to facilitate the multi-camera tracking component 6 to dynamically correct the matrix transformation relationship between the tracking coordinate system and the global coordinate system to achieve point cloud stitching. The operation cabinet is electrically connected to the robotic arm 5, the electric turntable 7, and the backlight marking points 3 on the measurement plane of the marble base, and contains an automated control computer terminal for controlling the movement of the robotic arm 5 and the electric turntable 7 and the brightness of the backlight marking points 3 on the measurement plane of the marble base. The CAD design model of the measurement target is preset and imported into the computer in the operation cabinet for subsequent deviation comparison with the measured point cloud data of the measurement target. The ceramic calibration plate is placed on the tabletop of the electric turntable 7 and is within the field of view of the point cloud reconstruction component 4, and is used to calibrate the point cloud reconstruction component 4 and dynamically solve the matrix transformation relationship between the scanning coordinate system of the point cloud reconstruction component 4 and the feature point coordinate system of the feature points on the outer shell of the point cloud reconstruction component 4 under the scanning path. The honeycomb aluminum calibration plate is placed within the field of view of the multi-camera tracking component 6 and is used to calibrate the multi-camera tracking component 6.

[0164] It should be noted that when calibrating the multi - camera tracking component 6 using the honeycomb aluminum calibration plate, the calibration of the multi - camera tracking component 6 is based on the principle of close - range photogrammetry. The honeycomb aluminum calibration plate is a calibration plate with dot - matrix arranged marking points that match the measurement field of view of the multi - camera tracking component 6. The essence of calibrating the multi - camera tracking component 6 is to use the global point coordinates of all marking points on the known honeycomb aluminum plate and the two - dimensional image point coordinates captured by the corresponding multi - cameras, and bundle - optimize and adjust the internal and external parameters of the cameras, so as to complete the calibration of the multi - camera tracking component 6. The automatic calibration of the point cloud reconstruction component 4 is carried out using the ceramic calibration plate. The automatic calibration of the point cloud reconstruction component 4 uses the adjustment optimization method. The ceramic calibration plate is a dot - matrix calibration plate with circular marking points regularly arranged. The point cloud reconstruction component 4 is fixed on the robotic arm 5. By moving the position of the ceramic calibration plate, the point cloud reconstruction component 4 collects a series of images of different positions of the calibration plate to achieve camera calibration. The matrix transformation relationship between the scanning coordinate system of the point cloud reconstruction component 4 and the feature point coordinate system of the feature points on the shell of the point cloud reconstruction component 4 is solved using the ceramic calibration plate. The ceramic calibration plate is placed on the electric turntable 7, and the ceramic calibration plate contains coded points and non - coded points. Ensure that the point cloud reconstruction component 4 and the multi - camera tracking component 6 can always simultaneously capture the ceramic calibration plate, and move the point cloud reconstruction component 4 to different positions within the space to be measured according to the planned solution path. A scanning coordinate system is established on the point cloud reconstruction component 4 in advance, a feature point coordinate system is established at the position of the marking points on the shell of the point cloud reconstruction component 4, a calibration plate coordinate system is established on the ceramic calibration plate, a tracking coordinate system is at the position of the multi - camera tracking component 6, and a global coordinate system is established near the backlight point of the marble measurement table. Through the ceramic calibration plate in this application, the matrix transformation relationship between the scanning coordinate system of the point cloud reconstruction component 4 and the feature point coordinate system of the feature points on the shell of the point cloud reconstruction component 4 is solved.

[0165] In this application, the multi - camera tracking component 6 simultaneously tracks the feature points on the shell of the point cloud reconstruction component 4 and the backlight marking points 3 on the marble measurement platform, dynamically corrects the matrix transformation relationship between the tracking coordinate system and the global coordinate system at each path point. At the same time, it dynamically solves the matrix transformation relationship between the scanning coordinate system and the tracking coordinate system at the same path point, and quickly converts the three - dimensional point cloud data of the measurement target measured by the point cloud reconstruction component 4 at each path point to the global coordinate system to achieve point cloud stitching. It dynamically corrects the accuracy loss problems caused by environmental vibration and temperature changes during measurement; the introduction of the electric turntable 7 reduces the spatial tracking range of the tracking component and increases the system freedom degree to 7 degrees, meeting the high - precision and high - efficiency measurement requirements for large - scale workpiece measurement, and ensuring the efficiency and accuracy of point cloud stitching.

[0166] As described above, these are only several embodiments of the present application and do not impose any form of limitation on the present application. Although the present application is disclosed above with preferred embodiments, it is not intended to limit the present application. Any person skilled in the art, without departing from the scope of the technical solution of the present application, making some changes or modifications using the technical content disclosed above is equivalent to equivalent implementation cases and all fall within the scope of the technical solution.

Claims

1. A multi-camera three-dimensional point cloud stitching method based on dynamic correction, characterized in that, Including: Step 1: Calibrate the multi - view tracking component (6), the point cloud reconstruction component (4), the feature points on the point cloud reconstruction component (4), and the measurement platform (1) where the measurement target (2) is located, and establish a tracking coordinate system, a scanning coordinate system, a feature point coordinate system, and a global coordinate system; Step 2: Move the point cloud reconstruction component (4) along the planned solution path to solve the matrix transformation relationship between the feature point coordinate system and the scanning coordinate system; Step 3: The point cloud reconstruction component (4) measures the measurement target (2) along the planned scanning path. The multi - view tracking component (6) tracks the feature points on the point cloud reconstruction component (4) and the backlight marking points (3) on the measurement platform (1), and solves the matrix transformation relationship between the tracking coordinate system and the global coordinate system and the matrix transformation relationship between the scanning coordinate system and the tracking coordinate system for each scanning path; In the said Step 3, dynamically solving the matrix transformation relationship between the tracking coordinate system and the global coordinate system and the matrix transformation relationship between the scanning coordinate system and the tracking coordinate system for each scanning path includes: Step 3.1: The multi - view tracking component (6) tracks the backlight marking points (3) on the measurement platform (1) at each scanning path point, and dynamically corrects the transformation relationship from the tracking coordinate system to the global coordinate system at the current path point; Suppose that the three-dimensional coordinates of the common marker point on the backlight marker point (3) in the tracking coordinate system obtained at the current scanning path point are P k_global_tarck , then: where R k_global and t k_global are the rotation matrix and translation matrix for transforming the three-dimensional coordinates of the backlight marker point (3) captured by the multi-view tracking component (6) at the k-th scanning path point in the global coordinate system to the tracking coordinate system; represents the matrix conversion relationship from the global coordinate system to the tracking coordinate system at the current scanning path point; P k_global is the three-dimensional coordinate of the corresponding common marker point in the global coordinate system; Step 3.2: The multi - view tracking component (6) simultaneously tracks the feature points on the point cloud reconstruction component (4) at the same scanning path point, and dynamically solves the transformation relationship from the feature point coordinate system to the tracking coordinate system of the feature points on the point cloud reconstruction component (4) at the current scanning path point; Suppose that the three-dimensional coordinates of the common feature points of the feature points in the tracking coordinate system obtained by the multi-object tracking component (6) under the current scanning path point are P k_target_track , then: Among them, and are the rotation matrix and translation matrix for transforming the three-dimensional coordinates of the k-th path point in the feature point coordinate system and the feature points captured by the multi-view tracking component (6) into the tracking coordinate system; represents the matrix transformation relationship from the feature point coordinate system to the tracking coordinate system under the current path point; P k_target is the three-dimensional coordinate of the feature points on the point cloud reconstruction component (4) captured by the multi-view tracking component (6) under the k-th path point in the feature point coordinate system; Step 4: The point cloud reconstruction component (4) simultaneously photographs the measurement target (2) and the feature points on the feature polyhedron (8), and converts the three - dimensional point cloud data of the measurement target (2) measured by the point cloud reconstruction component (4) into coordinate values in the same reference coordinate system; wherein, the measurement target (2) and the feature polyhedron (8) are both located on the electric turntable (7) of the measurement platform (1); Step 5: Convert the three - dimensional point cloud of the measurement target (2) and the coordinate data of the feature points of the feature polyhedron (8) in the scanning coordinate system corresponding to the current scanning path point scanned by the point cloud reconstruction component (4) to the point coordinate data in the scanning coordinate system corresponding to the first scanning path point, and convert them to the global coordinate system and complete point cloud stitching.

2. The multi-camera three-dimensional point cloud stitching method based on dynamic correction according to claim 1, wherein, The said Step 2 includes: Step 2.1: The point cloud reconstruction component (4) runs to the i - th solution path point, photographs the calibration plate and determines the transformation relationship between the scanning coordinate system and the calibration plate coordinate system; Step 2.2: The multi - view tracking coordinate system determines the transformation relationship between the tracking coordinate system and the feature point coordinate system by photographing the feature points on the point cloud reconstruction component (4); Step 2.3: Calculate the transformation relationship from the scanning coordinate system to the feature point coordinate system through the transformation relationship from the calibration plate coordinate system to the scanning coordinate system and the transformation relationship from the feature point coordinate system to the tracking coordinate system.

3. The method for multi-camera three-dimensional point cloud stitching based on dynamic correction according to claim 2, wherein Step 2.1 includes: Let the three-dimensional point coordinates of the common marker points on the calibration board in the scanning coordinate system captured at the i-th solution path point be denoted as p i_Scan , then: Among them, R i_base and t i_base respectively represent the rotation matrix and translation matrix from the calibration plate coordinate system to the scanning coordinate system under the i-th solution path point, and p i_base is the three-dimensional coordinate of the common marking point corresponding to p i_Scan in the calibration plate coordinate system; is the transformation matrix from the calibration plate coordinate system to the scanning coordinate system under the current solution path point.

4. The method for multi-camera three-dimensional point cloud stitching based on dynamic correction according to claim 2, wherein Step 2.2 includes: At the i-th solution path point, the three-dimensional coordinates of the feature points on the point cloud reconstruction component (4) in the tracking coordinate system obtained by the multi-object tracking component (6) are denoted as p i_track , then: Among them, i represents the number of solution path points, i ∈ {1, 2, 3, …, n}, R i_target and t i_target are respectively the rotation matrix and the translation matrix from the feature point coordinate system to the tracking coordinate system under the i-th solution path point; p i_target is the three-dimensional coordinate of the common feature point corresponding to p i_track in the feature point coordinate system; is the transformation matrix from the feature point coordinate system to the tracking coordinate system under the current solution path point.

5. The method for multi-camera three-dimensional point cloud stitching based on dynamic correction according to claim 3, wherein, Step 2.3 includes: Each solution path point satisfies: For all solution path points The matrix transformation relationship remains unchanged. Therefore, all n path points satisfy: Simplified representation: Ax = Bx; Where: Determine the matrix conversion relationship from the calibration board coordinate system to the tracking coordinate system at all solution path points Determine the matrix conversion relationship from the scanning coordinate system to the feature point coordinate system at all solution path points Among them, represents the inverse matrix of the matrix transformation relationship from the feature point coordinate system to the tracking coordinate system under the nth solution path; It represents the inverse matrix of the matrix transformation relationship from the calibration board coordinate system to the scanning coordinate system under the nth solution path.

6. The method for multi-camera three-dimensional point cloud stitching based on dynamic correction according to claim 1, wherein In step 4, the point cloud reconstruction component (4) simultaneously photographs the measurement target (2) and the feature points on the feature polyhedron (8), and converts the three-dimensional point cloud data of the measurement target (2) measured by the point cloud reconstruction component (4) into coordinate values in the same reference coordinate system, including: Step 4.1: The point cloud reconstruction component (4) captures the point cloud data P of the measurement target (2) and the feature polyhedron (8) with feature points in the scanning coordinate system corresponding to the first scanning path point. 1_Scan The point cloud reconstruction component (4) continues to move along the scanning path and captures the point cloud data P of the measurement target (2) and the feature polyhedron (8) with feature points in the scanning coordinate system corresponding to the current scanning path point. k_Scan Among them, there are at least 4 common feature points in the point cloud data obtained by scanning adjacent path points. Based on the common feature point coordinate data of the measurement target (2) scanned by the point cloud reconstruction component (4) at adjacent path points, solve the matrix transformation relationship between the two scanning coordinate systems, that is: Among them, k represents the number of points on the scanning path, k ∈ {1, 2, 3, …, m}, and p k_Scan is the coordinate data of the feature points on the feature polyhedron (8) in the scanning coordinate system corresponding to the current path point scanned by the point cloud reconstruction component (4) under the k-th scanning path point; p k-1_Scan is the coordinate data of the feature points on the corresponding feature polyhedron (8) in the scanning coordinate system corresponding to the current path point scanned by the point cloud reconstruction component (4) under the (k - 1)-th scanning path point; R k_Scan and t k_Scan are the rotation matrix and translation matrix for transforming the three-dimensional coordinates of the feature points on the feature polyhedron (8) in the scanning coordinate system corresponding to the k-th scanning path point to the scanning coordinate system corresponding to the (k - 1)-th scanning path point; represents the matrix conversion relationship between the k-th scanning coordinate system and the scanning coordinate system corresponding to the (k - 1)-th scanning path point under the current path point; Step 4.

2. At the k-th scanning path point, the matrix transformation relationship between the scanning coordinate system corresponding to the point cloud reconstruction component (4) and the scanning coordinate system corresponding to the first scanning path point is: Among them, p k_Scan is the coordinate data of the feature points on the feature polyhedron (8) in the scanning coordinate system corresponding to the current path point scanned by the point cloud reconstruction component (4) under the k-th scanning path point; p 1_Scan is the coordinate data of the feature points on the feature polyhedron (8) in the scanning coordinate system corresponding to the current path point scanned by the point cloud reconstruction component (4) under the first scanning path point; represents the matrix transformation relationship between the scanning coordinate system corresponding to the k-th scanning path point and the scanning coordinate system corresponding to the first path point; Determine the matrix transformation relationship between the three-dimensional point cloud data of the measurement target (2) photographed by the point cloud reconstruction component (4) and the feature point data on the feature polyhedron (8) at the k-th scanning path point in the scanning coordinate system corresponding to the first scanning path point: Among them, P k_scan_all is the three-dimensional point cloud of the measurement target (2) and the coordinate data of the feature points of the feature polyhedron (8) scanned by the point cloud reconstruction component (4) under the scanning coordinate system corresponding to the k-th scanning path point; P k_Scan_1 is the point coordinate data of the three-dimensional point cloud of the measurement target (2) and the coordinate data of the feature points of the feature polyhedron (8) under the scanning coordinate system corresponding to the k-th scanning path point in the scanning coordinate system corresponding to the first scanning path point.

7. The method for multi-camera three-dimensional point cloud stitching based on dynamic correction according to claim 1, wherein In step 5, convert to the global coordinate system and complete the point cloud stitching, including: Among them, P k_scan_global is the three-dimensional point cloud data of the measurement target (2) scanned by the point cloud reconstruction component (4) and the point coordinate data of the feature points of the feature polyhedron (8) in the global coordinate system under the k-th scanning path point; is the inverse matrix of the matrix transformation relationship from the global coordinate system to the tracking coordinate system under the k-th scanning path point; is the matrix transformation relationship from the feature point coordinate system to the tracking coordinate system under the k-th scanning path point; is the matrix transformation relationship from the scanning coordinate system to the feature point coordinate system during the measurement process. Since the scanning coordinate system of the point cloud reconstruction component (4) and the feature coordinate system of the feature points on the point cloud reconstruction component (4) are rigidly connected, the relative positions of these two coordinate systems remain fixed whether it is under the solution path or the scanning path, that is is a known quantity; P k_Scan_1 is the point coordinate data of the three-dimensional point cloud of the measurement target (2) and the feature point coordinates of the feature polyhedron (8) in the scanning coordinate system corresponding to the k-th scanning path point in the scanning coordinate system corresponding to the first scanning path point; There are at least 4 common feature points in the point cloud data of adjacent path points, and the point cloud stitching in the global coordinate system is realized through the common feature points.

8. The method for multi-camera three-dimensional point cloud stitching based on dynamic correction according to claim 1, wherein It also includes: Step 6. Package the three-dimensional point cloud data of the stitched measurement target (2) to obtain a complete mesh data model of the measurement target (2), and perform detection and comparison with the CAD design model corresponding to the measurement target (2).

9. The multi-camera three-dimensional measurement system based on dynamic correction used in the multi-camera three-dimensional point cloud stitching method based on dynamic correction according to claim 1, characterized in that, It includes: A measurement platform (1), on which a measurement target (2) and backlight marking points (3) are arranged; A point cloud reconstruction component (4), which is installed on the measurement platform (1) through a robotic arm (5), and the point cloud reconstruction component (4) measures the measurement target (2); A multi-camera tracking component (6), which performs tracking measurement on the measurement target (2) on the measurement platform (1), and feature points are arranged on the point cloud reconstruction component (4) and are within the field of view of the multi-camera tracking component (6); Among them, the measurement target (2) is installed on the measurement platform (1) through an electric turntable (7) and is located at the center of the electric turntable (7). There are also multiple feature polyhedrons (8) with feature points on the tabletop of the electric turntable (7), and the multiple feature polyhedrons (8) are located around the measurement target (2); The backlight marking points (3) are located on the periphery of the electric turntable (7).

Images