A point cloud data splicing method and device with progressive convergence error

Abstract

The application discloses a point cloud data splicing method and device with gradual convergence error, and the method comprises the following steps: obtaining corresponding point pairs in a first group of point cloud data and a second group of point cloud data according to a preset maximum response distance threshold; calculating a rigid transformation matrix according to the corresponding point pairs; performing rigid transformation on the first group of point cloud data, and calculating data variance of corresponding point pairs in the first group of point cloud data and the second group of point cloud data after transformation; if the data variance is greater than or equal to a preset variance threshold, reducing the preset maximum response distance threshold, recalculating the rigid transformation matrix and performing rigid transformation on the first group of point cloud data after transformation until the data variance is less than the preset variance threshold. By adopting the embodiments provided by the application, the maximum response distance is reduced for multiple times to ensure the stability of the point cloud data group after transformation through iteration of the rigid transformation matrix, and the emergence of noise points is avoided.

Description

Technical Field

[0001] This invention relates to the field of data processing technology, and in particular to a method and apparatus for stitching point cloud data with progressive convergence error. Background Technology

[0002] Point cloud registration can be done manually (relying on instruments) or automatically. Automatic point cloud registration uses algorithms or statistical rules to calculate the misalignment between two point cloud samples, achieving automatic registration. Essentially, it transforms the point cloud data measured in different coordinate systems to obtain a holistic data model. The key is determining the coordinate transformation parameters to minimize the distance between the 3D data measured from two different viewpoints after the transformation. Currently, registration algorithms can be categorized into global registration and local registration. PCL has a separate registration module that implements the basic data structures related to registration, and classic registration algorithms include ICP.

[0003] Current point cloud data stitching methods mainly involve coarse stitching using feature point matching, followed by fine stitching using ICP (point-to-point response distance), GICP, or ICP (point-to-plane response distance). However, because point cloud data acquired through real-world scanning contains noise, the fine stitching stage may result in incorrect matching when using ICP to find the nearest point, leading to unsatisfactory results. Summary of the Invention

[0004] This invention provides a point cloud data stitching method and apparatus with progressive convergence error reduction. It iterates the rigid transformation matrix by reducing the maximum response distance multiple times to ensure the stability of the transformed point cloud data set and avoid the occurrence of noise.

[0005] To achieve the above objectives, a first aspect of this application provides a point cloud data stitching method with progressively narrowing convergence error, comprising:

[0006] Based on the preset maximum response distance threshold, corresponding point pairs in the first set of point cloud data and the second set of point cloud data are obtained;

[0007] Calculate the rigid transformation matrix based on the corresponding point pairs;

[0008] A rigid transformation is performed on the first set of point cloud data, and the variance of the corresponding point pairs in the transformed first set of point cloud data and second set of point cloud data is calculated.

[0009] If the data variance is greater than or equal to a preset variance threshold, the preset maximum response distance threshold is reduced, the rigid transformation matrix is ​​recalculated, and the first set of point cloud data after transformation is subjected to rigid transformation until the data variance is less than the preset variance threshold.

[0010] If the data variance is less than a preset variance threshold, save the rigid transformation matrix and the first set of point cloud data at this time; the first set of point cloud data and the second set of point cloud data constitute a convergently stitched point cloud map.

[0011] In one possible implementation of the first aspect, obtaining corresponding point pairs from the first set of point cloud data and the second set of point cloud data according to a preset maximum response distance threshold specifically includes:

[0012] Extract the first set of key points and the second set of key points from the first set of point cloud data and the second set of point cloud data according to preset standards;

[0013] By combining the coordinate positions of the first set of key points in the first set of point cloud data and the second set of key points in the second set of point cloud data, the maximum response distance between the first set of key points and the second set of key points is calculated.

[0014] If the maximum response distance is less than the maximum response distance threshold, the first set of key points and the second set of key points form a corresponding point pair.

[0015] In one possible implementation of the first aspect, the rigid transformation matrix includes a rotation matrix and a translation vector.

[0016] In one possible implementation of the first aspect, the rigid transformation of the first set of point cloud data and the calculation of the data variance of corresponding point pairs in the transformed first set of point cloud data and second set of point cloud data specifically include:

[0017] Based on the covariance matrix formed by the first set of key points and the second set of key points, calculate the data variance of corresponding point pairs in the transformed first set of point cloud data and second set of point cloud data.

[0018] In one possible implementation of the first aspect, reducing the preset maximum response distance threshold specifically involves:

[0019] Set the preset maximum response distance threshold to half of its original value.

[0020] A second aspect of this application provides a point cloud data stitching device with progressively narrowing convergence error, comprising:

[0021] The point-to-point acquisition module is used to obtain corresponding point pairs in the first set of point cloud data and the second set of point cloud data according to the preset maximum response distance threshold.

[0022] The matrix calculation module is used to calculate the rigid transformation matrix based on the corresponding point pairs;

[0023] The variance calculation module is used to perform a rigid transformation on the first set of point cloud data and calculate the data variance of corresponding point pairs in the transformed first set of point cloud data and second set of point cloud data.

[0024] The progressive convergence module is used to reduce the preset maximum response distance threshold if the data variance is greater than or equal to a preset variance threshold, recalculate the rigid transformation matrix, and perform a rigid transformation on the transformed first set of point cloud data until the data variance is less than the preset variance threshold; if the data variance is less than the preset variance threshold, save the rigid transformation matrix and the first set of point cloud data at this time; the first set of point cloud data and the second set of point cloud data constitute a progressively converged and stitched point cloud map.

[0025] In one possible implementation of the second aspect, the point-to-point acquisition module is specifically used for:

[0026] Extract the first set of key points and the second set of key points from the first set of point cloud data and the second set of point cloud data according to preset standards;

[0027] By combining the coordinate positions of the first set of key points in the first set of point cloud data and the second set of key points in the second set of point cloud data, the maximum response distance between the first set of key points and the second set of key points is calculated.

[0028] If the maximum response distance is less than the maximum response distance threshold, the first set of key points and the second set of key points form a corresponding point pair.

[0029] In one possible implementation of the second aspect, the rigid transformation matrix includes a rotation matrix and a translation vector.

[0030] In one possible implementation of the second aspect, the step of rigidly transforming the first set of point cloud data and calculating the data variance of corresponding point pairs in the transformed first set of point cloud data and second set of point cloud data specifically includes: calculating the data variance of corresponding point pairs in the transformed first set of point cloud data and second set of point cloud data based on the covariance matrix formed by the first set of key points and the second set of key points.

[0031] In one possible implementation of the second aspect, reducing the preset maximum response distance threshold specifically involves:

[0032] Set the preset maximum response distance threshold to half of its original value.

[0033] Compared to existing technologies, the point cloud data stitching method and apparatus provided in this invention provides a progressive convergence error reduction method. By repeatedly setting the maximum response distance, the rigid transformation matrix is ​​iteratively calculated until the variance of the corresponding point pairs of the rigid transformation matrix and the transformed first and second sets of point cloud data is less than a preset threshold. This ensures that the rigid transformation of the first set of point cloud data using the rigid transformation matrix corresponding to the maximum response distance can cover more noise points, resulting in a more accurate stitching effect. This leads to a high overlap rate in the overlapping areas of the point clouds, significantly reducing the occurrence of seams in the stitching results in actual production. Attached Figure Description

[0034] Figure 1 This is a flowchart illustrating a point cloud data stitching method for progressively converging error provided in an embodiment of the present invention.

[0035] Figure 2 This is a schematic diagram of the first set of point cloud data and the second set of point cloud data before fine assembly in one embodiment of the present invention;

[0036] Figure 3 This is a schematic diagram of the first and second sets of point cloud data after fine assembly according to an embodiment of the present invention.

[0037] Figure 4 This is a point cloud image obtained by coarse stitching according to an embodiment of the present invention;

[0038] Figure 5 This is a result image of a point cloud image after fine stitching using a gradual convergence stitching method according to an embodiment of the present invention;

[0039] Figure 6 This is a partial view of the point cloud image before fine stitching in one embodiment of the present invention;

[0040] Figure 7 This is a partial image of a point cloud map after fine stitching using a gradual convergence stitching method in one embodiment of the present invention. Detailed Implementation

[0041] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0042] Please see Figure 1 The first aspect of this application provides a point cloud data stitching method with progressively narrowing convergence error, comprising:

[0043] S10. Based on the preset maximum response distance threshold, obtain the corresponding point pairs in the first set of point cloud data and the second set of point cloud data.

[0044] S11. Calculate the rigid transformation matrix based on the corresponding point pairs.

[0045] S12. Perform a rigid transformation on the first set of point cloud data, and calculate the variance of the corresponding point pairs in the transformed first set of point cloud data and second set of point cloud data.

[0046] S13. If the data variance is greater than or equal to a preset variance threshold, reduce the preset maximum response distance threshold, recalculate the rigid transformation matrix, and perform a rigid transformation on the first set of point cloud data after transformation until the data variance is less than the preset variance threshold.

[0047] S14. If the data variance is less than a preset variance threshold, save the rigid transformation matrix and the first set of point cloud data at this time; the first set of point cloud data and the second set of point cloud data constitute a convergently stitched point cloud map.

[0048] This invention primarily addresses the error problem caused by noise during the fine-stitching process through a progressively convergent stitching method, thereby ensuring the stitching result is as accurate as possible. Following the method provided in this embodiment, two sets of point cloud data are stitched together:

[0049] 1. Set CorrespondenceDistance (maximum response distance) to find the nearest corresponding point;

[0050] 2. Calculate the rigid transformation matrix based on the nearest corresponding point;

[0051] 3. Calculate the variance of the transformed data (also use CorrespondenceDistance to calculate the distance between selected points). If the variance is less than the threshold, the process is considered complete.

[0052] 4. Set CorrespondenceDistance to half of its original value (progression rate, can be modified);

[0053] 5. Calculate the rigid transformation matrix again for the transformed point cloud and calculate the variance of the transformed data (using the current CorrespondenceDistance). If the error is less than the previous one, continue to repeat steps 3, 4, and 5; if the error is greater than the previous one, stop.

[0054] For example, obtaining corresponding point pairs in the first set of point cloud data and the second set of point cloud data according to a preset maximum response distance threshold specifically includes:

[0055] Extract the first set of key points and the second set of key points from the first set of point cloud data and the second set of point cloud data according to preset standards;

[0056] By combining the coordinate positions of the first set of key points in the first set of point cloud data and the second set of key points in the second set of point cloud data, the maximum response distance between the first set of key points and the second set of key points is calculated.

[0057] If the maximum response distance is less than the maximum response distance threshold, the first set of key points and the second set of key points form a corresponding point pair.

[0058] See Figure 2 , Figure 2 This shows the status of the first two sets of point cloud data after fine assembly. Figure 3 This illustrates the state of the two sets of point cloud data after fine-tuning. In the diagram, points 1 and 1', 2 and 2', and 3 and 3' are three key point pairs in the two sets of point cloud data, forming three corresponding point pairs. Points 1, 2, and 3 are the key points of the first set of point cloud data, while points 1', 2', and 3' are the key points of the second set of point cloud data.

[0059] For example, the rigid transformation matrix includes a rotation matrix and a translation vector.

[0060] Two 3D points in three-dimensional space, Their Euclidean distance is expressed as: The goal of the 3D point cloud matching problem is to find the matrices R and T that transform P and Q, i.e., the parameters R (rotation matrix) and T (translation vector) of the coordinate transformation.

[0061] For example, the step of performing a rigid transformation on the first set of point cloud data and calculating the data variance of corresponding point pairs in the transformed first set of point cloud data and second set of point cloud data specifically includes:

[0062] Based on the covariance matrix formed by the first set of key points and the second set of key points, calculate the data variance of corresponding point pairs in the transformed first set of point cloud data and second set of point cloud data.

[0063] For example, reducing the preset maximum response distance threshold specifically involves:

[0064] Set the preset maximum response distance threshold to half of its original value.

[0065] Taking a single-point cloud map as an example, Figure 4 , Figure 6 These are schematic diagrams of the overall point cloud image and a partial view before stitching. Figure 5 , Figure 7 These are schematic diagrams of the overall and partial views of the stitched point cloud image. Especially from... Figure 7It is evident that using the method provided in this embodiment for splicing can significantly reduce the occurrence of seams in the splicing results (see...). Figure 6 The overlap rate of the point cloud overlapping area is high.

[0066] Compared to existing technologies, the point cloud data stitching method provided in this embodiment of the invention provides a progressive convergence error reduction method. By repeatedly setting the maximum response distance, the rigid transformation matrix is ​​iteratively calculated until the variance of the corresponding point pairs of the rigid transformation matrix and the transformed first and second sets of point cloud data is less than a preset threshold. This ensures that the rigid transformation of the first set of point cloud data using the rigid transformation matrix corresponding to the maximum response distance can cover more noise points, resulting in a more accurate stitching effect. This leads to a high overlap rate in the overlapping areas of the point clouds, which greatly reduces the occurrence of seams in the stitching results in actual production.

[0067] A second aspect of this application provides a point cloud data stitching device with progressive convergence error, comprising: a point pair acquisition module, a matrix calculation module, a variance calculation module, and a progressive convergence module.

[0068] The point-to-point acquisition module is used to obtain corresponding point pairs in the first set of point cloud data and the second set of point cloud data according to the preset maximum response distance threshold.

[0069] The matrix calculation module is used to calculate the rigid transformation matrix based on the corresponding point pairs;

[0070] The variance calculation module is used to perform a rigid transformation on the first set of point cloud data and calculate the data variance of corresponding point pairs in the transformed first set of point cloud data and second set of point cloud data.

[0071] The progressive convergence module is used to reduce the preset maximum response distance threshold if the data variance is greater than or equal to a preset variance threshold, recalculate the rigid transformation matrix, and perform a rigid transformation on the transformed first set of point cloud data until the data variance is less than the preset variance threshold; if the data variance is less than the preset variance threshold, save the rigid transformation matrix and the first set of point cloud data at this time; the first set of point cloud data and the second set of point cloud data constitute a progressively converged and stitched point cloud map.

[0072] For example, the point-to-point acquisition module is specifically used for:

[0073] Extract the first set of key points and the second set of key points from the first set of point cloud data and the second set of point cloud data according to preset standards;

[0074] By combining the coordinate positions of the first set of key points in the first set of point cloud data and the second set of key points in the second set of point cloud data, the maximum response distance between the first set of key points and the second set of key points is calculated.

[0075] If the maximum response distance is less than the maximum response distance threshold, the first set of key points and the second set of key points form a corresponding point pair.

[0076] For example, the rigid transformation matrix includes a rotation matrix and a translation vector.

[0077] For example, the step of performing a rigid transformation on the first set of point cloud data and calculating the data variance of corresponding point pairs in the transformed first set of point cloud data and second set of point cloud data specifically includes: calculating the data variance of corresponding point pairs in the transformed first set of point cloud data and second set of point cloud data based on the covariance matrix formed by the first set of key points and the second set of key points.

[0078] For example, reducing the preset maximum response distance threshold specifically involves:

[0079] Set the preset maximum response distance threshold to half of its original value.

[0080] Compared to existing technologies, the point cloud data stitching device provided in this embodiment of the invention provides a progressive convergence error reduction method. By repeatedly setting the maximum response distance, the rigid transformation matrix is ​​iteratively calculated until the variance of the corresponding point pairs of the rigid transformation matrix and the transformed first and second sets of point cloud data is less than a preset threshold. This ensures that the rigid transformation of the first set of point cloud data using the rigid transformation matrix corresponding to the maximum response distance can cover more noise points, resulting in a more accurate stitching effect. This leads to a high overlap rate in the overlapping areas of the point clouds, which greatly reduces the occurrence of seams in the stitching results in actual production.

[0081] Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the specific working process of the device described above can be referred to the corresponding process in the foregoing method embodiments, and will not be elaborated further here.

[0082] The above description represents the preferred embodiments of the present invention. It should be noted that those skilled in the art can make various improvements and modifications without departing from the principles of the present invention, and these improvements and modifications are also considered to be within the scope of protection of the present invention.

Claims

1. A point cloud data stitching method with progressively narrowing convergence error, characterized in that, include: Based on the preset maximum response distance threshold, corresponding point pairs are obtained from the first set of point cloud data and the second set of point cloud data; Calculate the rigid transformation matrix based on the corresponding point pairs; A rigid transformation is performed on the first set of point cloud data, and the variance of the corresponding point pairs in the transformed first set of point cloud data and second set of point cloud data is calculated. If the data variance is greater than or equal to a preset variance threshold, the preset maximum response distance threshold is reduced, the rigid transformation matrix is ​​recalculated, and the first set of point cloud data after transformation is subjected to rigid transformation until the data variance is less than the preset variance threshold. If the data variance is less than a preset variance threshold, save the rigid transformation matrix and the first set of point cloud data at this time; the first set of point cloud data and the second set of point cloud data constitute a convergently stitched point cloud map. The reduction of the preset maximum response distance threshold specifically involves setting the preset maximum response distance threshold to half of its original value.

2. The point cloud data stitching method with progressive convergence error as described in claim 1, characterized in that, The step of obtaining corresponding point pairs in the first set of point cloud data and the second set of point cloud data according to a preset maximum response distance threshold specifically includes: Extract the first set of key points and the second set of key points from the first set of point cloud data and the second set of point cloud data according to preset standards; By combining the coordinate positions of the first set of key points in the first set of point cloud data and the second set of key points in the second set of point cloud data, the maximum response distance between the first set of key points and the second set of key points is calculated. If the maximum response distance is less than the maximum response distance threshold, the first set of key points and the second set of key points form a corresponding point pair.

3. The point cloud data stitching method with progressively narrowing convergence error as described in claim 1, characterized in that, The rigid transformation matrix includes a rotation matrix and a translation vector.

4. The point cloud data stitching method with progressive convergence error as described in claim 2, characterized in that, The step of performing a rigid transformation on the first set of point cloud data and calculating the variance of corresponding point pairs in the transformed first set of point cloud data and second set of point cloud data specifically includes: Based on the covariance matrix formed by the first set of key points and the second set of key points, calculate the data variance of corresponding point pairs in the transformed first set of point cloud data and second set of point cloud data.

5. A point cloud data stitching device with progressive convergence error reduction, characterized in that, include: The point-to-point acquisition module is used to obtain corresponding point pairs in the first set of point cloud data and the second set of point cloud data according to the preset maximum response distance threshold. The matrix calculation module is used to calculate the rigid transformation matrix based on the corresponding point pairs; The variance calculation module is used to perform a rigid transformation on the first set of point cloud data and calculate the data variance of corresponding point pairs in the transformed first set of point cloud data and second set of point cloud data. The progressive convergence module is used to reduce the preset maximum response distance threshold if the data variance is greater than or equal to a preset variance threshold, recalculate the rigid transformation matrix, and perform a rigid transformation on the transformed first set of point cloud data until the data variance is less than the preset variance threshold; if the data variance is less than the preset variance threshold, save the rigid transformation matrix and the first set of point cloud data at this time; the first set of point cloud data and the second set of point cloud data constitute the progressively converged and stitched point cloud map. The reduction of the preset maximum response distance threshold specifically involves setting the preset maximum response distance threshold to half of its original value.

6. The point cloud data stitching device with progressive convergence error as described in claim 5, characterized in that, The point-to-point acquisition module is specifically used for: Extract the first set of key points and the second set of key points from the first set of point cloud data and the second set of point cloud data according to preset standards; By combining the coordinate positions of the first set of key points in the first set of point cloud data and the second set of key points in the second set of point cloud data, the maximum response distance between the first set of key points and the second set of key points is calculated. If the maximum response distance is less than the maximum response distance threshold, the first set of key points and the second set of key points form a corresponding point pair.

7. The point cloud data stitching device with progressive convergence error as described in claim 5, characterized in that, The rigid transformation matrix includes a rotation matrix and a translation vector.

8. The point cloud data stitching device with progressive convergence error as described in claim 6, characterized in that, The step of performing a rigid transformation on the first set of point cloud data and calculating the variance of corresponding point pairs in the transformed first set of point cloud data and second set of point cloud data specifically includes: Based on the covariance matrix formed by the first set of key points and the second set of key points, calculate the data variance of corresponding point pairs in the transformed first set of point cloud data and second set of point cloud data.

Images