Point cloud registration method and device combining real and virtual target spots, and storage medium

Abstract

The invention belongs to the technical field of earth surface deformation monitoring, and discloses a point cloud registration method and device combining real and virtual target spots, and a storage medium, and the method comprises the steps: laying an entity target ball in a near shore, selecting an artificial building to construct a virtual target spot, and finally combining the entity target ball and the virtual target spot to construct a target spot matrix of a long baseline. According to the method, sparse point clouds of remotely-arranged entity targets are avoided through near-shore entity target balls, the problem of cross-river arrangement is avoided by constructing virtual target spots, and the three bottlenecks of target failure, error amplification and feature scarcity in point cloud registration after cross-river ultra-long-distance scanning of a ground laser radar are broken through; the method is suitable for landslide monitoring and river-crossing long-distance monitoring, and is especially suitable for surface deformation monitoring of large-span natural scenes such as reservoirs in high and steep mountainous areas, rivers along banks and the like.

Description

Technical Field

[0001] This invention belongs to the field of surface deformation monitoring technology, specifically relating to a point cloud registration method for joint real and virtual target points, which is particularly suitable for surface deformation monitoring in large-span natural scenes such as reservoirs in high and steep mountainous areas and riverbanks. It also relates to a computer device and storage medium. Background Technology

[0002] Terrestrial three-dimensional laser scanning (TLS) is a method for rapidly acquiring high-density, high-precision three-dimensional coordinates of the Earth's surface. It can observe different phases of the deformed area of ​​the Earth's surface and, after a series of post-processing steps, obtain a high-resolution deformation field of the entire deformed area. In this process, the error of point cloud registration will directly affect the final calculated deformation accuracy. In cross-river ultra-long-distance surface deformation monitoring, the terrestrial lidar needs to acquire the complete monitoring area point cloud by deploying equipment on the opposite bank. The existing point cloud registration technology has the following defects in this scenario: (1) Due to the influence of the laser divergence angle, the distance between scanning points at ultra-long distances increases to more than 10 cm, while the diameter of a conventional target sphere does not exceed 20 cm. This will result in sparse point clouds on the surface of the target sphere with a smaller diameter, making it impossible to fit the center coordinates of the sphere. If a super-large target sphere with a diameter greater than 1 m is deployed, the cost will be high and the implementation will be difficult; (2) If the target sphere is only deployed about 100 m from the shore near the scanner, while the monitoring area is located at 1.5–2 The opposite bank of km will cause the registration error to be significantly amplified due to the influence of the error propagation law and the short side controlling the long side; (3) Stable rigid features are scarce in natural landforms, and the traditional automatic registration method based on point, line and surface features has a high failure rate. Summary of the Invention

[0003] The purpose of this invention is to address the aforementioned problems in the prior art by providing a point cloud registration method for joint real and virtual target points, as well as a computer device and storage medium.

[0004] The above-mentioned objectives of the present invention are achieved by the following technical means:

[0005] A point cloud registration method for joint real and virtual target points includes the following steps:

[0006] Step 1: Obtain the target study area The scan data of each physical target sphere is used to calculate the center coordinates of each target sphere. The sphere number is... The coordinates of the center of the solid target sphere are , The number of physical target balls. The serial number of the physical target ball. Take 1~ ;

[0007] Step 2: Construct from scan data A virtual target and obtain The coordinates of a virtual target point;

[0008] Step 3, The center of the target sphere and A target matrix is ​​constructed by jointly identifying virtual target points. ;

[0009] Step 4: Repeat steps 1-3 to obtain the target point matrix of multiple periods of point cloud data. Select one period of point cloud data as the reference point cloud data, register the other point cloud data to the reference point cloud data, and calculate the rotation matrix and translation vector of the other period point cloud data to the reference point cloud data.

[0010] Step 5: Based on the rotation matrix and translation vector of the other point cloud data from Step 4 to the reference point cloud data, register the other point cloud data to the reference point cloud data.

[0011] As described above, step 2 specifically includes the following steps:

[0012] Step 2.1: Select non-natural buildings that do not undergo deformation from the target study area, segment out the point cloud of three non-parallel planes, and calculate the coordinates of the intersection of the three non-parallel planes as the coordinates of the virtual target point;

[0013] Step 2.2: Repeat step 2.1 to construct... The virtual target points, of which the first one is the first virtual target point. The coordinates of the virtual target points are , The number of virtual target points. The index of the virtual target. Take 1~ .

[0014] The target matrix in step 3 as described above Based on the following formula:

[0015] ;

[0016] In the formula, For the serial number The coordinates of the center of the solid target sphere. For the serial number The coordinates of the center of the solid target sphere. The coordinates of the first virtual target point. For the first The coordinates of a virtual target.

[0017] The coordinates of the intersection point of the three non-parallel planes in step 2.1 above are calculated as follows:

[0018] First, the point clouds of the three non-parallel planes are fitted to plane equations using the least squares method:

[0019] ;

[0020] ;

[0021] In the formula, To find the minimum value function, It is the first The plane equations and normal vectors of the planes are respectively in direction, direction, and Directional component vectors, For plane number, Take 1~3, This is a constant offset. For the first The number of planar point clouds in a plane. The point cloud sequence number is used in step 2.1. The sequence number is the point cloud in step 2.1. For the serial number The coordinates of the planar point cloud, These are constraints;

[0022] Secondly, by solving the system of equations for the three non-parallel planes, the coordinates of the intersection point of the three non-parallel planes, which is the virtual target point, are calculated. :

[0023] ;

[0024] In the formula, It is the inverse matrix. The plane equation and normal vector of the first plane are respectively in direction, direction, and Directional component vectors, The plane equation and normal vector of the second plane are respectively in direction, direction, and Directional component vectors, The normal vectors of the plane equations of the third plane are respectively in direction, direction, and Directional component vectors, All are constant offsets.

[0025] As mentioned above, step 1, calculating the center coordinates of each target sphere, specifically includes the following steps:

[0026] Step 1.2.1: Cut out the spherical point cloud from the scan data of the solid target sphere;

[0027] Step 1.2.2: The coordinates of the sphere's center are calculated using the least squares method based on the following formula:

[0028] ;

[0029] In the formula, To find the minimum value function, Let the coordinates be the center coordinates of the solid target sphere. Let the radius of the solid target sphere be . The point cloud sequence number is used in step 1.2.2. The index of the spherical point cloud of the solid target sphere. The number of spherical point clouds of the solid target sphere, in step 1.2.2 For the serial number The coordinates of the spherical point cloud of the solid target sphere;

[0030] Step 1.2.3: Repeat steps 1.2.1 and 1.2.2 to calculate the center coordinates of each target sphere in turn.

[0031] As described above, step 4 specifically includes the following steps:

[0032] Step 4.1: Repeat steps 1-3 to obtain the target point matrix of M periods of point cloud data. Select one period of point cloud data as the baseline point cloud data, and designate the baseline point cloud data as the first period of point cloud data. The target point matrix of the point cloud data is , For point cloud scanning periods, Take values ​​from 1 to M, where M is the total number of point cloud scanning periods;

[0033] Step 4.2: Calculate the center of the target point matrix of the point cloud data in each period, and center the target point matrix of the point cloud data in each period to obtain the centered matrix of the point cloud data in each period.

[0034] Among them, the Target Matrix Center of Point Cloud Data Calculated based on the following formula:

[0035] ;

[0036] In the formula, It is the index of the physical or virtual target in the target matrix. Take 1~ , This represents the total number of physical and virtual targets in the target matrix.

[0037] No. Centralized matrix of point cloud data Calculated based on the following formula:

[0038] ;

[0039] Step 4.3: Calculate the covariance matrix of other point cloud data to the baseline point cloud data based on the following formula:

[0040] ;

[0041] In the formula, For the first The covariance matrix from the initial point cloud data to the baseline point cloud data. A centralized matrix of baseline point cloud data, For the first Centralized matrix of point cloud data The transpose of the matrix;

[0042] Step 4.4: Perform singular value decomposition on each covariance matrix based on the following formula:

[0043] ;

[0044] In the formula, For the first The left singular vector matrix of the point cloud data. For the first The singular value matrix of the point cloud data. For the first The right singular vector matrix of the point cloud data. For the first The transpose of the right singular vector matrix of the point cloud data; and Satisfies orthogonality: , , It is the identity matrix. It is a left singular vector matrix. It is a right singular vector matrix;

[0045] Step 4.5: Calculate the rotation matrix and translation vector of the other point cloud data registered to the reference point cloud data based on the following formulas:

[0046] ;

[0047] ;

[0048] In the formula, For the first Rotation matrix for registering reference point cloud data to base point cloud data. For the first The translation vector of the initial point cloud data is registered to the reference point cloud data. For the first The transpose of the left singular vector matrix of the point cloud data. The target point matrix center is the reference point cloud data.

[0049] As described above, in step 5, other point cloud data are registered to the reference point cloud data based on the following formula:

[0050] ;

[0051] In the formula, For the registered first Point cloud data; For the first time before registration Point cloud data; For the first The transpose of the rotation matrix of the point cloud data.

[0052] A computer device includes a memory and a processor, the memory storing a computer program, and the processor executing the computer program to implement the steps of a point cloud registration method for joint real and virtual target points as described above.

[0053] A computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the steps of a point cloud registration method for joint real and virtual target points as described above.

[0054] Compared with the prior art, the present invention has the following advantages:

[0055] (1) Overcoming distance limitations: This invention is applicable to landslide monitoring and long-distance monitoring across rivers, especially to surface deformation monitoring in large-span natural scenes such as reservoirs in high and steep mountainous areas and riverbanks. This invention deploys physical target spheres near the shore, which can avoid the problem of sparse point cloud when deploying physical targets at long distances; and then constructs virtual target points to avoid the problem of deployment across rivers. The method of this invention overcomes the three major bottlenecks of target failure, error amplification and feature scarcity in point cloud registration after ultra-long-distance scanning across rivers by ground lidar, and significantly improves registration accuracy and reliability.

[0056] (2) Suppressing error propagation: This invention combines near-shore physical target spheres and far-distance virtual target points to form a long control baseline (scanning from near-shore to monitoring far-shore), which reduces registration error;

[0057] (3) Improved robustness: The present invention constructs a plane on an artificial building. The stability of the artificial building plane is better than that of natural features, and the virtual target point can be repeatedly constructed. Attached Figure Description

[0058] Figure 1 This is a flowchart of the method of the present invention;

[0059] Figure 2 This is a schematic diagram of scanning data for an example of a physical target sphere according to the present invention;

[0060] Figure 3 For the present invention from Figure 1 The spherical point cloud map cut out from the box in the image;

[0061] Figure 4 This is a schematic diagram illustrating an example of the present invention of constructing virtual target points by selecting non-natural buildings that do not undergo deformation from the monitoring area;

[0062] Figure 5 For the present invention from Figure 4 A schematic diagram of the planar point cloud segmented within the box;

[0063] Figure 6 This is a diagram showing the effect of the first corner point after registration using the registration method that only uses a solid target ball in Embodiment 2 of the present invention;

[0064] Figure 7 This is a diagram showing the effect of the second corner point after registration using a registration method that only uses a solid target ball in Embodiment 2 of the present invention;

[0065] Figure 8 This is an image of the first corner point after registration using the method of manual point selection for coarse registration followed by nearest neighbor iteration (ICP) for fine registration in Embodiment 2 of the present invention.

[0066] Figure 9 This is an image of the second corner point after registration using the method of manual point selection for coarse registration followed by nearest neighbor iteration (ICP) for fine registration in Embodiment 2 of the present invention.

[0067] Figure 10 This is a diagram showing the effect of the first corner point after registration using the method of the present invention in Embodiment 2 of the present invention;

[0068] Figure 11 This is a diagram showing the effect of the second corner point after registration using the method of the present invention in Embodiment 2 of the present invention. Detailed Implementation

[0069] To facilitate understanding and implementation of the present invention by those skilled in the art, the present invention will be further described in detail below with reference to embodiments. The embodiments described herein are for illustration and explanation only and are not intended to limit the present invention.

[0070] Example 1:

[0071] like Figure 1As shown, a point cloud registration method for joint real and virtual target points specifically includes the following steps:

[0072] In this embodiment, scanning data from three physical target spheres and two virtual target points are used, with the three-dimensional directions being respectively... direction, direction, direction.

[0073] Step 1, as follows Figure 2 and Figure 3 As shown, scan data of three physical target spheres in the target study area were acquired, and the center coordinates of each target sphere were obtained from the scan data. The scan data is from a ground-based lidar scanner (such as a Riegl VZ2000 model) that was deployed to the area of ​​surface deformation to collect the data. The specific steps include:

[0074] Step 1.1: Three physical target spheres are placed on the same side of the scanner. The diameter of the physical target spheres is 14cm and the distance from the scanner does not exceed 100m. The spherical point cloud of the three physical target spheres is obtained by scanning.

[0075] Step 1.2: Calculate the coordinates of the centers of the three target spheres, specifically including the following steps:

[0076] Step 1.2.1: Use CloudCompare software (a 3D point cloud processing software) to cut out the spherical point cloud from the scan data of the solid target sphere.

[0077] Step 1.2.2: Let the coordinates of the center of the solid target sphere be... Any point in a spherical point cloud The sphere's equation (i.e., formula (1)) is satisfied. Then, the least squares method is used to fit the sphere's equation, and the coordinates of the sphere's center are calculated:

[0078] (1);

[0079] (2);

[0080] In the formula, To find the minimum value function, Let the radius of the solid target sphere be . The point cloud sequence number is used in step 1.2.2. The index of the spherical point cloud of the solid target sphere. The number of spherical point clouds representing the solid target sphere, as described in step 1.2.2. For the serial number The coordinates of the spherical point cloud of the solid target sphere.

[0081] Step 1.2.3: Repeat steps 1.2.1 and 1.2.2 to calculate the center coordinates of each target sphere in turn, where the sequence number is... The coordinates of the center of the solid target sphere are , The serial number of the physical target ball. Take 1~ , The number of physical target balls.

[0082] Step 2: Construct two virtual target points from the scan data and obtain the coordinates of the two virtual target points. This includes the following steps:

[0083] Step 2.1, as follows Figure 4 and Figure 5 As shown, using CloudCompare software, non-natural structures (such as houses, bridge piers, tunnel entrances, anti-slide piles, etc.) that do not undergo deformation are selected from the monitoring area. Three non-parallel planes are segmented into point clouds (each plane has more than 200 point clouds). The coordinates of the intersection of the three non-parallel planes are then calculated as the coordinates of the virtual target point. The specific steps include:

[0084] Step 2.1.1, the first The planar point cloud dataset of the plane is as follows: The plane equation is given by formula (3). The plane point cloud dataset of non-parallel planes is fitted to the plane equation by the least squares method:

[0085] (3);

[0086] (4);

[0087] (5);

[0088] In the formula, It is the first Plane equations and normal vectors of a plane In respectively direction, direction, and The direction component vector, the first The normal vector of the plane equation of the nth plane is perpendicular to the nth plane. The first plane is used to control the second plane. Spatial directions of a plane; This is a constant offset, representing the translation distance of the control plane along the normal vector direction; For plane number, Take 1 to 3; The point cloud sequence number is used in step 2.1. The index is the number of the planar point cloud. For the first The number of planar point clouds in each plane, in step 2.1 For the serial number The coordinates of the planar point cloud; These are constraints.

[0089] Step 2.1.2: Solve the system of plane equations simultaneously and calculate the coordinates of the intersection point of the three planes as the virtual target point. :

[0090] (6);

[0091] In the formula, It is the inverse matrix. The plane equation and normal vector of the first plane are respectively in direction, direction, and Directional component vectors, The plane equation and normal vector of the second plane are respectively in direction, direction, and Directional component vectors, The normal vectors of the plane equations of the third plane are respectively in direction, direction, and Directional component vectors, All are constant offsets.

[0092] This invention selects to construct a plane on an artificial building, and the stability of the artificial building plane is superior to that of natural features.

[0093] Step 2.2: Repeat step 2.1 to construct two virtual target points, where the first... The coordinates of the virtual target points are , The number of virtual target points. The index of the virtual target. Take 1~ In this embodiment =2.

[0094] Step 3: Construct a target matrix by combining the center of the physical target sphere with the virtual target points. :

[0095] (7);

[0096] In the formula, These are the coordinates of the center of the target sphere with serial number 1. These are the coordinates of the center of the target sphere with serial number 2. These are the coordinates of the center of the target sphere with serial number 3. These are the coordinates of the first virtual target. These are the coordinates of the second virtual target.

[0097] Step 3 of the present invention constructs a target point matrix with a long baseline by combining the center of the physical target and the virtual target points, thereby forming a long control baseline, reducing registration error, and avoiding the point cloud sparsity of physical targets deployed at long distances and the registration error of targets deployed across rivers.

[0098] Step 4: Repeat steps 1-3 to obtain the target point matrix of multiple point cloud data. Select one period of point cloud data as the reference point cloud data, register other point cloud data to the reference point cloud data, and calculate the rotation matrix and translation vector of other point cloud data to the reference point cloud data. The specific steps include:

[0099] Step 4.1: Repeat steps 1-3 to obtain the target point matrix of M periods of point cloud data. Select one period of point cloud data as the baseline point cloud data, and designate the baseline point cloud data as the first period of point cloud data. The target point matrix of the point cloud data is , For point cloud scanning periods, Take values ​​from 1 to M, where M is the total number of point cloud scanning periods.

[0100] Step 4.2: Calculate the target point matrix center of the point cloud data for each period, and center the target point matrix of the point cloud data for each period to obtain the centered matrix of the point cloud data for each period, specifically:

[0101] (8);

[0102] (9);

[0103] In the formula, For the first The center of the target point matrix in the point cloud data. For the first Centralized matrix of point cloud data, It is the index of the physical or virtual target in the target matrix. Take 1~ , In this embodiment, the total number of physical targets and virtual targets in the target matrix is ​​[number]. =5.

[0104] Step 4.3: Calculate the covariance matrix of other point cloud data to the baseline point cloud data, based on the following formula:

[0105] (10);

[0106] In the formula, For the first The covariance matrix from the initial point cloud data to the baseline point cloud data. A centralized matrix of baseline point cloud data, For the first Centralized matrix of point cloud data The transpose of .

[0107] Step 4.4: Perform singular value decomposition (SVD) on each covariance matrix obtained in Step 4.3, based on the following formula:

[0108] (11);

[0109] In the formula, For the first The left singular vector matrix of the point cloud data is such that its column vectors are orthogonal and spanned by a covariance matrix. The line space; For the first The singular value matrix of the point cloud data, the diagonal elements of the singular value matrix (arranged in descending order) represent the variance contribution in each direction; For the first The right singular vector matrix of the point cloud data is such that its column vectors are orthogonal and spanned by a covariance matrix. The column space; For the first The transpose of the right singular vector matrix of the period; and Satisfies orthogonality: , ( It is the identity matrix. It is a left singular vector matrix. (This is a right singular vector matrix).

[0110] Step 4.5: Calculate the rotation matrix and translation vector of the other point cloud data registered to the reference point cloud data. The rotation matrix and translation vector are calculated based on the following formulas:

[0111] (12);

[0112] (13);

[0113] In the formula, For the first Rotation matrix for registering reference point cloud data to base point cloud data. For the first The translation vector of the initial point cloud data is registered to the reference point cloud data. For the first The transpose of the left singular vector matrix of the point cloud data; The target point matrix center is the reference point cloud data.

[0114] Step 5: Based on the rotation matrix and translation vector of the other point cloud data from Step 4 to the reference point cloud data, register the other point cloud data to the reference point cloud data, based on the following formula:

[0115] (14);

[0116] In the formula, For the registered first Point cloud data; For the first time before registration Point cloud data; For the first The transpose of the rotation matrix of the point cloud data.

[0117] The method of this invention overcomes the three major bottlenecks in point cloud registration after long-distance scanning across rivers by ground-based lidar: target failure, error amplification, and feature scarcity. It significantly improves registration accuracy and reliability, and is suitable for landslide monitoring and long-distance monitoring across rivers. It is especially suitable for monitoring surface deformation in large-span natural scenes such as reservoirs in high and steep mountainous areas and riverbanks.

[0118] Example 2:

[0119] This embodiment was tested at the Dawanzi landslide and Tuandigou deformation body located in the headwaters of a hydropower station. In this scenario, the river surface is wide after the reservoir is filled. Due to the limitation of the vertical field of view of the ground three-dimensional laser scanning TLS, ultra-long-distance scanning across the river is unavoidable. Therefore, a ground observation network was set up on the opposite bank at a distance of about 1.5 to 2.3 km from the observation target, and two phases of ground point cloud data were collected in July and December 2024, respectively.

[0120] This embodiment employs registration methods using only physical target spheres (such as...). Figure 6 and Figure 7 As shown), the method of manually selecting points for coarse registration and then using nearest neighbor iteration (ICP) for fine registration (e.g.) Figure 8 and Figure 9 (as shown), and the method of the present invention (as shown). Figure 10 and Figure 11After registration (as shown), two specific corner points were selected for detailed evaluation. The traditional method using only target spheres resulted in errors of 0.87m and 0.72m at these two specific corner points. The method using manual point selection for coarse registration followed by nearest-neighbor iteration for fine registration resulted in errors of 5.5m and 3.81m at these two specific corner points. The method of this invention scanned three nearby physical target spheres and constructed two virtual target points near the observation area on the opposite bank, ultimately achieving errors of 0.11m and 0.13m at these two specific corner points. In comparison, from... Figures 6-11 It can be seen that the method of the present invention performs better in point cloud registration in ultra-long-distance cross-river scenarios.

[0121] In one embodiment, a computer device is also provided, including a memory and a processor, wherein the memory stores a computer program, and the processor executes the computer program to implement the steps in the above method embodiments.

[0122] In one embodiment, a computer-readable storage medium is provided having a computer program stored thereon that, when executed by a processor, implements the steps in the above method embodiments.

[0123] In one embodiment, a computer program product is provided, including a computer program that, when executed by a processor, implements the steps in the above method embodiments.

[0124] It should be noted that the embodiments described in this invention are merely illustrative of the spirit of the invention. Those skilled in the art to which this invention pertains can make various modifications or additions to the described embodiments or use similar methods to substitute them, without departing from the spirit of the invention or exceeding the scope defined by the appended claims.

Claims

1. A point cloud registration method for joint real and virtual target points, characterized in that, Includes the following steps: Step 1: Obtain the target study area The scan data of each physical target sphere is used to calculate the center coordinates of each target sphere. The sphere number is... The coordinates of the center of the solid target sphere are , The number of physical target balls. The serial number of the physical target ball. Take 1~ ; Step 2: Construct from scan data A virtual target and obtain The coordinates of a virtual target point; Step 3, The center of the target sphere and A target matrix is ​​constructed by jointly identifying virtual target points. ; Step 4: Repeat steps 1-3 to obtain the target point matrix of multiple periods of point cloud data. Select one period of point cloud data as the reference point cloud data, register the other point cloud data to the reference point cloud data, and calculate the rotation matrix and translation vector of the other period point cloud data to the reference point cloud data. Step 5: Based on the rotation matrix and translation vector of the other point cloud data from Step 4 to the reference point cloud data, register the other point cloud data to the reference point cloud data.

2. The point cloud registration method for joint real and virtual target points according to claim 1, characterized in that, Step 2 specifically includes the following steps: Step 2.1: Select non-natural buildings that do not undergo deformation from the target study area, segment out the point cloud of three non-parallel planes, and calculate the coordinates of the intersection of the three non-parallel planes as the coordinates of the virtual target point; Step 2.2: Repeat step 2.1 to construct... The virtual target points, of which the first one is the first virtual target point. The coordinates of the virtual target points are , The number of virtual target points. The index of the virtual target. Take 1~ .

3. The point cloud registration method for joint real and virtual target points according to claim 2, characterized in that, Target matrix in step 3 Based on the following formula: ； In the formula, For the serial number The coordinates of the center of the solid target sphere. For the serial number The coordinates of the center of the solid target sphere. The coordinates of the first virtual target point. For the first The coordinates of a virtual target.

4. The point cloud registration method for joint real and virtual target points according to claim 2, characterized in that, The coordinates of the intersection point of the three non-parallel planes in step 2.1 are calculated in the following way: First, the point clouds of the three non-parallel planes are fitted to plane equations using the least squares method: ； ； In the formula, To find the minimum value function, It is the first The plane equations and normal vectors of the planes are respectively in direction, direction, and Directional component vectors, For plane number, Take 1~3, This is a constant offset. For the first The number of planar point clouds in a plane. The point cloud sequence number is used in step 2.

1. The index of the planar point cloud is used in step 2.

1. For the serial number The coordinates of the planar point cloud, These are constraints; Secondly, by solving the system of equations for the three non-parallel planes, the coordinates of the intersection point of the three non-parallel planes, which is the virtual target point, are calculated. : ； In the formula, It is the inverse matrix. The plane equation and normal vector of the first plane are respectively in direction, direction, and Directional component vectors, The plane equation and normal vector of the second plane are respectively in direction, direction, and Directional component vectors, The normal vectors of the plane equations of the third plane are respectively in direction, direction, and Directional component vectors, All are constant offsets.

5. The point cloud registration method for joint real and virtual target points according to claim 1, characterized in that, The calculation of the center coordinates of each target sphere in step 1 specifically includes the following steps: Step 1.2.1: Cut out the spherical point cloud from the scan data of the solid target sphere; Step 1.2.2: The coordinates of the sphere's center are calculated using the least squares method based on the following formula: ； In the formula, To find the minimum value function, Let the coordinates be the center coordinates of the solid target sphere. Let the radius of the solid target sphere be . The point cloud sequence number is used in step 1.2.

2. The index of the spherical point cloud of the solid target sphere. The number of spherical point clouds of the solid target sphere, in step 1.2.2 For the serial number The coordinates of the spherical point cloud of the solid target sphere; Step 1.2.3: Repeat steps 1.2.1 and 1.2.2 to calculate the center coordinates of each target sphere in turn.

6. The point cloud registration method for joint real and virtual target points according to claim 4, characterized in that, Step 4 specifically includes the following steps: Step 4.1: Repeat steps 1-3 to obtain the target point matrix of M periods of point cloud data. Select one period of point cloud data as the baseline point cloud data, and designate the baseline point cloud data as the first period of point cloud data. The target point matrix of the point cloud data is , For point cloud scanning periods, Take values ​​from 1 to M, where M is the total number of point cloud scanning periods; Step 4.2: Calculate the center of the target point matrix of the point cloud data in each period, and center the target point matrix of the point cloud data in each period to obtain the centered matrix of the point cloud data in each period. Among them, the Target Matrix Center of Point Cloud Data Calculated based on the following formula: ； In the formula, It is the index of the physical or virtual target in the target matrix. Take 1~ , This represents the total number of physical and virtual targets in the target matrix. No. Centralized matrix of point cloud data Calculated based on the following formula: ； Step 4.3: Calculate the covariance matrix of other point cloud data to the baseline point cloud data based on the following formula: ； In the formula, For the first The covariance matrix from the initial point cloud data to the baseline point cloud data. A centralized matrix of baseline point cloud data, For the first Centralized matrix of point cloud data The transpose of the matrix; Step 4.4: Perform singular value decomposition on each covariance matrix based on the following formula: ； In the formula, For the first The left singular vector matrix of the point cloud data. For the first The singular value matrix of the point cloud data. For the first The right singular vector matrix of the point cloud data. For the first The transpose of the right singular vector matrix of the point cloud data; and Satisfies orthogonality: , , It is the identity matrix. It is a left singular vector matrix. It is a right singular vector matrix; Step 4.5: Calculate the rotation matrix and translation vector of the other point cloud data registered to the reference point cloud data based on the following formulas: ； ； In the formula, For the first Rotation matrix for registering reference point cloud data to base point cloud data. For the first The translation vector of the initial point cloud data is registered to the reference point cloud data. For the first The transpose of the left singular vector matrix of the point cloud data. The target point matrix center is the reference point cloud data.

7. The point cloud registration method for joint real and virtual target points according to claim 6, characterized in that, In step 5, other point cloud data are registered to the reference point cloud data based on the following formula: ； In the formula, For the registered first Point cloud data; For the first time before registration Point cloud data; For the first The transpose of the rotation matrix of the point cloud data.

8. A computer device comprising a memory and a processor, wherein the memory stores a computer program, characterized in that, When the processor executes the computer program, it implements the steps of the point cloud registration method for joint real and virtual target points as described in any one of claims 1 to 7.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by the processor, it implements the steps of the point cloud registration method for joint real and virtual target points as described in any one of claims 1 to 7.

Images