Point cloud seed point selection, point cloud clustering method and device
By projecting the point cloud onto the parameter space using the Hough transform and selecting seed points based on voxel density, the instability and high computational complexity caused by random seed point selection in point cloud clustering algorithms are solved, achieving efficient and stable point cloud clustering.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- ALIBABA INNOVATION PRIVATE LIMITED
- Filing Date
- 2021-06-17
- Publication Date
- 2026-06-19
AI Technical Summary
In existing point cloud clustering algorithms, the random selection of seed points leads to unstable clustering results, high computational complexity, and susceptibility to noise interference, especially on point clouds with small surface areas.
The point cloud is projected into the parameter space by Hough transform, and specific voxels are selected as seed points by using voxel density to avoid random selection and reduce the computational complexity to O(n).
It improves the stability and computational efficiency of point cloud clustering, reduces the amount of computation, and avoids the impact of noise on normal vector calculation.
Smart Images
Figure CN115496929B_ABST
Abstract
Description
Technical Field
[0001] This disclosure relates to the field of computer technology, specifically to point cloud seed point selection, point cloud clustering, and sensor data processing methods and apparatus. Background Technology
[0002] Point clouds are a data form that represents three-dimensional objects using dense three-dimensional points (x, y, z). For example, LiDAR is a common sensor in autonomous driving and high-precision map acquisition, and point clouds are the data representation of LiDAR. Point cloud clustering is a fundamental processing method in the processing and analysis of point cloud data. The purpose of point cloud clustering is to group point cloud points with the same properties / attributes into one class, and point cloud points with different properties / attributes into different classes. Related point cloud clustering algorithms include region growing algorithms, connected region analysis, K-means, and other growth-type (region growing, surface growing) clustering algorithms. Growth-type clustering algorithms are a type of clustering algorithm that clusters by selecting seed points and pre-defined growth rules. Seed points are selected randomly or by calculating normal vectors. The selection of seed points determines the quality of the final clustering.
[0003] The disadvantages of randomly selecting seed points are: strong randomness, inability to guarantee clustering effect, and different results when running the same data repeatedly, making it impossible to guarantee the consistency of results.
[0004] The disadvantages of selecting seed points by calculating normal vectors are: the computational cost of calculating normal vectors is large (the algorithm complexity is O(nlogn)), and it is easily affected by noise. It also results in unstable performance for clustering point clouds with small surface areas. Summary of the Invention
[0005] To address the problems in the related technologies, embodiments of this disclosure provide a point cloud seed point selection method and apparatus, a point cloud clustering method, a sensor data processing method and apparatus, an electronic device, a readable storage medium, and a computer program product.
[0006] In a first aspect, this disclosure provides a point cloud seed point selection method, including:
[0007] The point cloud is projected onto the parameter space according to the preset parameter model through a preset transformation method to obtain the voxel corresponding to the point cloud in the parameter space.
[0008] Specific voxels are selected from the voxels in the parameter space according to preset voxel selection rules;
[0009] Based on the correspondence between the point cloud and the voxels in the parameter space, points corresponding to the specific voxels are selected from the point cloud as seed points for point cloud clustering.
[0010] In conjunction with the first aspect, in the first implementation of the first aspect of this disclosure, the preset transformation method is the Hough transform.
[0011] In conjunction with the first implementation of the first aspect, in the second implementation of the first aspect of this disclosure, the step of projecting the point cloud onto the parameter space according to a preset parameter model through a preset transformation method to obtain the voxels corresponding to the point cloud in the parameter space includes:
[0012] The point cloud is projected into the parameter space through the Hough transform according to the preset planar model to obtain the voxels corresponding to the point cloud in the Hough parameter space.
[0013] In conjunction with the first aspect and any of the first to second implementations of the first aspect, in the third implementation of the first aspect of this disclosure, the step of selecting a specific voxel from the voxels in the parameter space according to a preset voxel selection rule includes:
[0014] Determine the density of the voxels corresponding to the unclustered points in the point cloud in the parameter space;
[0015] Select the voxel with the highest density from the voxels as the specific voxel corresponding to the unclustered point in the point cloud.
[0016] Secondly, this disclosure provides a point cloud seed point selection device, including:
[0017] The transformation module is configured to project the point cloud onto the parameter space using a preset transformation method according to a preset parameter model to obtain the voxel corresponding to the point cloud in the parameter space.
[0018] The voxel selection module is configured to select a specific voxel from the voxels in the parameter space according to a preset voxel selection rule;
[0019] The seed point selection module is configured to select points corresponding to a specific voxel from the point cloud as seed points for point cloud clustering based on the correspondence between the point cloud and the voxels in the parameter space.
[0020] In conjunction with the second aspect, in the first implementation of the second aspect of this disclosure, the voxel selection module includes:
[0021] The determination submodule is configured to determine the density of voxels corresponding to the unclustered points in the point cloud in the parameter space;
[0022] The voxel selection submodule is configured to select the voxel with the highest density from the voxels as the specific voxel corresponding to the unclustered point in the point cloud.
[0023] Thirdly, this disclosure provides a point cloud clustering method, including:
[0024] The point cloud is projected onto the parameter space according to the preset parameter model through a preset transformation method to obtain the voxel corresponding to the point cloud in the parameter space.
[0025] Specific voxels are selected from the voxels in the parameter space according to preset voxel selection rules;
[0026] Based on the correspondence between the point cloud and the voxels in the parameter space, points corresponding to the specific voxels are selected from the point cloud as seed points for point cloud clustering.
[0027] Point cloud clustering is performed using the seed points.
[0028] Fourthly, this disclosure provides an electronic device, including a memory and a processor; wherein,
[0029] The memory is used to store one or more computer instructions, wherein the one or more computer instructions are executed by the processor to implement the method as described in the first aspect, the first to the third implementation of the first aspect, and any one of the third aspects.
[0030] Fifthly, this disclosure provides a readable storage medium storing computer instructions that, when executed by a processor, implement the method described in the first aspect, the first to the third implementation of the first aspect, or any one of the third aspects.
[0031] In a sixth aspect, this disclosure provides a computer program product including computer instructions that, when executed by a processor, implement the method described in any one of the first aspect, the first to third implementations of the first aspect, and the third aspect.
[0032] The technical solutions provided in this disclosure may have the following beneficial effects:
[0033] According to the technical solution provided in this disclosure, a point cloud is projected onto a parameter space using a preset transformation method based on a preset parameter model to obtain the voxels corresponding to the point cloud in the parameter space; a specific voxel is selected from the voxels in the parameter space according to a preset voxel selection rule; and points corresponding to the specific voxels are selected from the point cloud as seed points for point cloud clustering based on the correspondence between the point cloud and the voxels in the parameter space. This avoids the problem of poor clustering results caused by randomly selecting seed points. Furthermore, it avoids calculating normal vectors, thus avoiding the calculation of feature matrices and neighborhood searches. Therefore, the computational complexity can be reduced from O(nlogn) to O(n), significantly reducing the computational load. Simultaneously, it avoids the problem of noise affecting normal vector calculation.
[0034] According to the technical solution provided in this disclosure, the preset transformation method is the Hough transform. Using the 3D Hough transform, point cloud points are projected onto the Hough space, and positions meeting certain conditions are selected as seed positions, thereby determining the corresponding seed points in the point cloud. This avoids the problem of poor clustering results caused by randomly selecting seed points. Furthermore, it eliminates the need to calculate normal vectors, thus avoiding the calculation of the feature matrix and neighborhood search. Therefore, the computational complexity can be reduced from O(nlogn) to O(n), significantly reducing the computational load. Simultaneously, it avoids the problem of noise affecting normal vector calculation.
[0035] According to the technical solution provided in this disclosure, the method of projecting a point cloud onto a parameter space using a preset transformation method based on a preset parameter model to obtain the voxels corresponding to the point cloud in the parameter space includes: projecting the point cloud onto the parameter space using a Hough transform based on a preset planar model to obtain the voxels corresponding to the point cloud in the Hough parameter space. A 3D Hough transform can be used to project the point cloud points onto the Hough space, selecting the positions that meet the conditions as seed positions to determine the corresponding seed points in the point cloud, thereby avoiding the problem of poor clustering results caused by randomly selecting seed points. Furthermore, it avoids calculating normal vectors, thus avoiding the calculation of the feature matrix and the need for neighborhood search. Therefore, the computational complexity can be reduced from O(nlogn) to O(n), significantly reducing the computational load. Simultaneously, it avoids the problem of noise affecting the calculation of normal vectors.
[0036] According to the technical solution provided in this disclosure, selecting specific voxels from voxels in the parameter space according to a preset voxel selection rule includes: determining the density of voxels corresponding to un-clustered points in the point cloud in the parameter space; selecting the voxel with the highest density from the voxels as the specific voxel corresponding to the un-clustered points in the point cloud. This can be achieved using a 3D Hough transform to project the point cloud points onto the Hough space. By determining the density of un-clustered points in the Hough space, the positions that meet the conditions are selected as seed positions, thereby determining the corresponding seed points in the point cloud. This avoids the problem of poor clustering results caused by randomly selecting seed points. Furthermore, it avoids calculating normal vectors, thus avoiding the calculation of the feature matrix and neighborhood search. Therefore, the computational complexity can be reduced from O(nlogn) to O(n), significantly reducing the computational load. Simultaneously, it avoids the problem of noise affecting normal vector calculation.
[0037] According to the technical solution provided in this disclosure, a transformation module is configured to project a point cloud onto a parameter space using a preset transformation method based on a preset parameter model to obtain voxels corresponding to the point cloud in the parameter space; a voxel selection module is configured to select specific voxels from the voxels in the parameter space according to preset voxel selection rules; and a seed point selection module is configured to select points corresponding to the specific voxels from the point cloud as seed points for point cloud clustering based on the correspondence between the point cloud and the voxels in the parameter space, thus avoiding the problem of poor clustering results caused by randomly selecting seed points. Furthermore, it avoids calculating normal vectors, thereby avoiding the calculation of feature matrices and neighborhood searches. Therefore, the computational complexity can be reduced from O(nlogn) to O(n), significantly reducing the computational load. Simultaneously, it avoids the problem of noise affecting normal vector calculation.
[0038] According to the technical solution provided in this disclosure, the voxel selection module includes: a determination submodule configured to determine the density of voxels corresponding to the un-clustered points in the point cloud in the parameter space; and a voxel selection submodule configured to select the voxel with the highest density from the voxels as the specific voxel corresponding to the un-clustered points in the point cloud. This can be achieved by using a 3D Hough transform to project the point cloud points onto the Hough space. By determining the density of the un-clustered points in the Hough space, the positions that meet the conditions are selected as seed positions, thereby determining the corresponding seed points in the point cloud. This avoids the problem of poor clustering results caused by randomly selecting seed points. Furthermore, it avoids calculating normal vectors, thus avoiding the calculation of the feature matrix and neighborhood search. Therefore, the computational complexity can be reduced from O(nlogn) to O(n), significantly reducing the computational load. Simultaneously, it avoids the problem of noise affecting normal vector calculation.
[0039] According to the technical solution provided in this disclosure, a point cloud is projected onto a parameter space through a preset transformation method based on a preset parameter model to obtain the voxels corresponding to the point cloud in the parameter space; a specific voxel is selected from the voxels in the parameter space according to a preset voxel selection rule; based on the correspondence between the point cloud and the voxels in the parameter space, a point corresponding to the specific voxel is selected from the point cloud as a seed point for point cloud clustering; point cloud clustering is performed using the seed point, avoiding the problem of poor clustering effect caused by randomly selecting seed points. Moreover, it is not necessary to calculate the normal vector, which avoids the calculation of the feature matrix and the neighborhood search. Therefore, the computational complexity of O(nlogn) can be reduced to O(n), significantly reducing the amount of computation. At the same time, it avoids the problem of normal vector calculation being affected by noise.
[0040] It should be understood that the above general description and the following detailed description are exemplary and explanatory only, and are not intended to limit this disclosure. Attached Figure Description
[0041] Other features, objects, and advantages of this disclosure will become more apparent from the following detailed description of non-limiting embodiments, taken in conjunction with the accompanying drawings. In the drawings:
[0042] Figure 1 A flowchart is shown illustrating a point cloud seed point selection method according to an embodiment of the present disclosure;
[0043] Figure 2 Implementation shown Figure 1 A flowchart illustrating a specific implementation of the point cloud seed point selection method;
[0044] Figure 3 Showing according to Figure 2 An exemplary schematic diagram of the Hough transform method in the point cloud seed point selection method is shown;
[0045] Figure 4 A structural block diagram of a point cloud seed point selection device according to an embodiment of the present disclosure is shown.
[0046] Figure 5 A flowchart illustrating a point cloud clustering method according to an embodiment of the present disclosure is shown;
[0047] Figure 6 A flowchart illustrating a sensor data processing method according to an embodiment of the present disclosure is shown;
[0048] Figure 7 A structural block diagram of a sensor data processing apparatus according to an embodiment of the present disclosure is shown;
[0049] Figure 8A structural block diagram of an electronic device according to an embodiment of the present disclosure is shown;
[0050] Figure 9 This is a schematic diagram of the structure of a computer system suitable for implementing the methods according to the embodiments of this disclosure. Detailed Implementation
[0051] In the following, exemplary embodiments of the present disclosure will be described in detail with reference to the accompanying drawings to enable those skilled in the art to readily implement them. Furthermore, for clarity, portions unrelated to the description of the exemplary embodiments have been omitted from the drawings.
[0052] In this disclosure, it should be understood that terms such as “comprising” or “having” are intended to indicate the presence of labels, numbers, steps, actions, components, parts or combinations thereof disclosed in this specification, and are not intended to exclude the possibility of one or more other labels, numbers, steps, actions, components, parts or combinations thereof being present or added.
[0053] It should also be noted that, unless otherwise specified, the embodiments and labels in this disclosure can be combined with each other. This disclosure will now be described in detail with reference to the accompanying drawings and embodiments.
[0054] According to embodiments of this disclosure, a novel method for selecting seed points is provided: seed points are selected through transformations such as 3D Hough transform, avoiding the problem of poor clustering results caused by random seed point selection, and eliminating the need to calculate normal vectors. This avoids the calculation of feature matrices and neighborhood search. The computational complexity is reduced from O(nlogn) to O(n), significantly reducing the computational load. Furthermore, it avoids the problem of noise affecting normal vector calculation.
[0055] According to the technical solution provided in this disclosure, a point cloud is projected onto a parameter space using a preset transformation method based on a preset parameter model to obtain the voxels corresponding to the point cloud in the parameter space; a specific voxel is selected from the voxels in the parameter space according to a preset voxel selection rule; and points corresponding to the specific voxels are selected from the point cloud as seed points for point cloud clustering based on the correspondence between the point cloud and the voxels in the parameter space. This avoids the problem of poor clustering results caused by randomly selecting seed points. Furthermore, it avoids calculating normal vectors, thus avoiding the calculation of feature matrices and neighborhood searches. Therefore, the computational complexity can be reduced from O(nlogn) to O(n), significantly reducing the computational load. Simultaneously, it avoids the problem of noise affecting normal vector calculation.
[0056] To address the aforementioned issues, this disclosure proposes a point cloud seed point selection method and apparatus, a point cloud clustering method, a sensor data processing method and apparatus, an electronic device, a readable storage medium, and a computer program product.
[0057] Figure 1 A flowchart illustrating a point cloud seed point selection method according to an embodiment of the present disclosure is shown. Figure 1 The point cloud seed point selection method shown includes steps S101, S102, and S103.
[0058] In step S101, the point cloud is projected onto the parameter space through a preset transformation method according to the preset parameter model to obtain the voxels corresponding to the point cloud in the parameter space.
[0059] In step S102, a specific voxel is selected from the voxels in the parameter space according to a preset voxel selection rule.
[0060] In step S103, based on the correspondence between the point cloud and the voxels in the parameter space, points corresponding to the specific voxels are selected from the point cloud as seed points for point cloud clustering.
[0061] In one embodiment of this disclosure, selecting seed points for a point cloud through normal vector analysis is computationally too intensive. Therefore, converting the point cloud data to a parameter space allows for seed point selection with lower computational cost. In one embodiment of this disclosure, a voxel is short for volume pixel, which is the smallest unit of digital data segmentation in three-dimensional space. Points in the point cloud can be projected onto the parameter space to obtain the corresponding voxels.
[0062] In one embodiment of this disclosure, the preset voxel selection rule may be to select the voxel in the densest region according to the voxel density, or to select the voxel according to a specific attribute of the voxel, etc. Those skilled in the art can understand how to select a specific voxel in the parameter space according to the preset voxel selection rule based on the teachings of the embodiments of this disclosure.
[0063] In one embodiment of this disclosure, since voxels are obtained by projecting points in a point cloud onto a parameter space through a preset transformation method, there is a correspondence between points in the point cloud and voxels in the parameter space. Therefore, by determining a specific voxel in the parameter space, the corresponding point in the point cloud can also be determined.
[0064] The following is for reference Figure 2 Description of implementation Figure 1 The flowchart illustrates a specific implementation of the point cloud seed point selection method. Figure 2 A flowchart is shown for a point cloud seed point selection method according to another embodiment of the present disclosure.
[0065] like Figure 2 As shown, the point cloud is first projected onto the parameter space using the Hough transform based on the parametric model M to obtain the voxel V in the Hough parameter space. The coordinates of point P in the point cloud are (x, y, z). The parametric model M can be, for example, a planar model:
[0066] ,
[0067] in, , , .
[0068] In one embodiment of this disclosure, the Hough Transform refers to a 3D Hough Transform, which can transform data from three-dimensional space to Hough parameter space based on a preset parameter model. See below for further details. Figure 3 describe Figure 2 The Hough transform in [the context of the original text]. Figure 3 Showing according to Figure 2 The diagram shows an exemplary schematic of the Hough transform method in the point cloud seed point selection method.
[0069] like Figure 3 As shown, a point P(x,y,z) in the point cloud can be projected into the Hough parameter space through a Hough transform. The point P(x,y,z) can be transformed into P(θ,ω,r) with parameters θ, ω, and r as variables. Here, θ is the angle between the line containing point P and the origin of the 3D Cartesian coordinate system and the Z-axis; ω is the angle between the projection of the line containing point P and the origin of the Cartesian coordinate system onto the XY plane Cartesian coordinate system and the X-axis; and r is the distance between point P and the origin of the Cartesian coordinate system. The parameters θ, ω, and r of the Hough parameter space can be used to define the voxel V corresponding to the point in the point cloud within the Hough parameter space.
[0070] In one embodiment of this disclosure, the preset transformation method is the Hough transform.
[0071] According to the technical solution provided in this disclosure, the preset transformation method is the Hough transform. Using the 3D Hough transform, point cloud points are projected onto the Hough space, and positions meeting certain conditions are selected as seed positions, thereby determining the corresponding seed points in the point cloud. This avoids the problem of poor clustering results caused by randomly selecting seed points. Furthermore, it eliminates the need to calculate normal vectors, thus avoiding the calculation of the feature matrix and neighborhood search. Therefore, the computational complexity can be reduced from O(nlogn) to O(n), significantly reducing the computational load. Simultaneously, it avoids the problem of noise affecting normal vector calculation.
[0072] In one embodiment of this disclosure, Figure 1 Step S101 includes: projecting the point cloud onto the parameter space through Hough transform according to the preset planar model to obtain the voxels corresponding to the point cloud in the Hough parameter space.
[0073] According to the technical solution provided in this disclosure, the method of projecting a point cloud onto a parameter space using a preset transformation method based on a preset parameter model to obtain the voxels corresponding to the point cloud in the parameter space includes: projecting the point cloud onto the parameter space using a Hough transform based on a preset planar model to obtain the voxels corresponding to the point cloud in the Hough parameter space. A 3D Hough transform can be used to project the point cloud points onto the Hough space, selecting the positions that meet the conditions as seed positions to determine the corresponding seed points in the point cloud, thereby avoiding the problem of poor clustering results caused by randomly selecting seed points. Furthermore, it avoids calculating normal vectors, thus avoiding the calculation of the feature matrix and the need for neighborhood search. Therefore, the computational complexity can be reduced from O(nlogn) to O(n), significantly reducing the computational load. Simultaneously, it avoids the problem of noise affecting the calculation of normal vectors.
[0074] like Figure 2 As shown, a specific voxel V' can be selected from voxel V in the Hough parameter space according to the voxel selection rules of the Hough parameter space.
[0075] In one embodiment of this disclosure, the voxel selection rule of the Hough parameter space may be to select the voxel in the densest region according to the voxel density of the Hough parameter space, or to select the voxel according to a specific attribute of the voxel, etc. Those skilled in the art can understand how to select a specific voxel in the Hough parameter space according to the voxel selection rule of the Hough parameter space based on the teachings of the embodiments of this disclosure.
[0076] In one embodiment of this disclosure, the preset voxel selection rule is to select the voxel with the highest density among the voxels as a specific voxel.
[0077] According to the technical solution provided in this disclosure, by using the preset voxel selection rule to select the voxel with the highest density among the voxels as specific voxels, a 3D Hough transform can be used to project the point cloud points onto the Hough space. By determining the density of un-clustered points in the Hough space, the positions that meet the conditions are selected as seed positions, thereby determining the corresponding seed points in the point cloud. This avoids the problem of poor clustering results caused by randomly selecting seed points. Moreover, it avoids the need to calculate normal vectors, thus avoiding the calculation of the feature matrix and the need for neighborhood search. Therefore, the computational complexity can be reduced from O(nlogn) to O(n), significantly reducing the amount of computation. At the same time, it avoids the problem of normal vector calculation being affected by noise.
[0078] like Figure 2 As shown, point P' corresponding to a specific voxel V' is selected as the seed point in the point cloud.
[0079] In one embodiment of this disclosure, Figure 1 Step S102 includes: determining the density of voxels corresponding to the un-clustered points in the point cloud in the parameter space; and selecting the voxel with the highest density from the voxels as the specific voxel corresponding to the un-clustered points in the point cloud.
[0080] According to the technical solution provided in this disclosure, selecting specific voxels from voxels in the parameter space according to a preset voxel selection rule includes: determining the density of voxels corresponding to un-clustered points in the point cloud in the parameter space; selecting the voxel with the highest density from the voxels as the specific voxel corresponding to the un-clustered points in the point cloud. This can be achieved using a 3D Hough transform to project the point cloud points onto the Hough space. By determining the density of un-clustered points in the Hough space, the positions that meet the conditions are selected as seed positions, thereby determining the corresponding seed points in the point cloud. This avoids the problem of poor clustering results caused by randomly selecting seed points. Furthermore, it avoids calculating normal vectors, thus avoiding the calculation of the feature matrix and neighborhood search. Therefore, the computational complexity can be reduced from O(nlogn) to O(n), significantly reducing the computational load. Simultaneously, it avoids the problem of noise affecting normal vector calculation.
[0081] In one embodiment of this disclosure, points in the point cloud are projected onto the Hough parameter space using a 3D Hough transform. By determining the density of voxels corresponding to un-clustered points in the Hough parameter space, the region containing the point corresponding to the densest voxel is selected as the seed region. Therefore, the point with the highest density can be selected as the seed point. In one embodiment of this disclosure, a statistical method can be used to determine the density of voxels corresponding to un-clustered points in the Hough parameter space.
[0082] The following is for reference Figure 4 This invention describes a point cloud seed point selection device according to an embodiment of the present disclosure. Figure 4 A structural block diagram of a point cloud seed point selection device 400 according to an embodiment of the present disclosure is shown.
[0083] like Figure 4 As shown, the point cloud seed point selection device 400 includes: a transformation module 401, a voxel selection module 402, and a seed point selection module 403.
[0084] The transformation module 401 is configured to project the point cloud onto the parameter space through a preset transformation method according to the preset parameter model to obtain the voxel corresponding to the point cloud in the parameter space.
[0085] The voxel selection module 402 is configured to select a specific voxel from the voxels in the parameter space according to a preset voxel selection rule.
[0086] The seed point selection module 403 is configured to select points corresponding to a specific voxel from the point cloud as seed points for point cloud clustering based on the correspondence between the point cloud and the voxels in the parameter space.
[0087] According to the technical solution provided in this disclosure, a transformation module is configured to project a point cloud onto a parameter space using a preset transformation method based on a preset parameter model to obtain voxels corresponding to the point cloud in the parameter space; a voxel selection module is configured to select specific voxels from the voxels in the parameter space according to preset voxel selection rules; and a seed point selection module is configured to select points corresponding to the specific voxels from the point cloud as seed points for point cloud clustering based on the correspondence between the point cloud and the voxels in the parameter space, thus avoiding the problem of poor clustering results caused by randomly selecting seed points. Furthermore, it avoids calculating normal vectors, thereby avoiding the calculation of feature matrices and neighborhood searches. Therefore, the computational complexity can be reduced from O(nlogn) to O(n), significantly reducing the computational load. Simultaneously, it avoids the problem of noise affecting normal vector calculation.
[0088] In one embodiment of this disclosure, the preset transformation method is the Hough transform.
[0089] According to the technical solution provided in this disclosure, the preset transformation method is the Hough transform. Using the 3D Hough transform, point cloud points are projected onto the Hough space, and positions meeting certain conditions are selected as seed positions, thereby determining the corresponding seed points in the point cloud. This avoids the problem of poor clustering results caused by randomly selecting seed points. Furthermore, it eliminates the need to calculate normal vectors, thus avoiding the calculation of the feature matrix and neighborhood search. Therefore, the computational complexity can be reduced from O(nlogn) to O(n), significantly reducing the computational load. Simultaneously, it avoids the problem of noise affecting normal vector calculation.
[0090] In one embodiment of this disclosure, the transformation module 401 is further configured to:
[0091] The point cloud is projected into the parameter space through the Hough transform according to the preset planar model to obtain the voxels corresponding to the point cloud in the Hough parameter space.
[0092] According to the technical solution provided in this disclosure, the transformation module is further configured to: project the point cloud onto the parameter space using a Hough transform based on a preset planar model to obtain the voxels corresponding to the point cloud in the Hough parameter space. A 3D Hough transform can be used to project the point cloud points onto the Hough space, selecting the positions that meet the conditions as seed positions to determine the corresponding seed points in the point cloud. This avoids the problem of poor clustering results caused by randomly selecting seed points. Furthermore, it eliminates the need to calculate normal vectors, thus avoiding the calculation of the feature matrix and neighborhood search. Therefore, the computational complexity can be reduced from O(nlogn) to O(n), significantly reducing the computational load. Simultaneously, it avoids the problem of noise affecting normal vector calculation.
[0093] In one embodiment of this disclosure, the preset voxel selection rule is to select the voxel with the highest density among the voxels as a specific voxel.
[0094] According to the technical solution provided in this disclosure, by using the preset voxel selection rule to select the voxel with the highest density among the voxels as specific voxels, a 3D Hough transform can be used to project the point cloud points onto the Hough space. By determining the density of un-clustered points in the Hough space, the positions that meet the conditions are selected as seed positions, thereby determining the corresponding seed points in the point cloud. This avoids the problem of poor clustering results caused by randomly selecting seed points. Moreover, it avoids the need to calculate normal vectors, thus avoiding the calculation of the feature matrix and the need for neighborhood search. Therefore, the computational complexity can be reduced from O(nlogn) to O(n), significantly reducing the amount of computation. At the same time, it avoids the problem of normal vector calculation being affected by noise.
[0095] In one embodiment of this disclosure, the voxel selection module 402 includes:
[0096] The determination submodule is configured to determine the density of voxels corresponding to the unclustered points in the point cloud in the parameter space;
[0097] The voxel selection submodule is configured to select the voxel with the highest density from the voxels as the specific voxel corresponding to the unclustered point in the point cloud.
[0098] According to the technical solution provided in this disclosure, the voxel selection module includes: a determination submodule configured to determine the density of voxels corresponding to the un-clustered points in the point cloud in the parameter space; and a voxel selection submodule configured to select the voxel with the highest density from the voxels as the specific voxel corresponding to the un-clustered points in the point cloud. This can be achieved by using a 3D Hough transform to project the point cloud points onto the Hough space. By determining the density of the un-clustered points in the Hough space, the positions that meet the conditions are selected as seed positions, thereby determining the corresponding seed points in the point cloud. This avoids the problem of poor clustering results caused by randomly selecting seed points. Furthermore, it avoids calculating normal vectors, thus avoiding the calculation of the feature matrix and neighborhood search. Therefore, the computational complexity can be reduced from O(nlogn) to O(n), significantly reducing the computational load. Simultaneously, it avoids the problem of noise affecting normal vector calculation.
[0099] Those skilled in the art will understand that, with reference to Figure 4 The described technical solution can be compared with the reference. Figures 1 to 3 The described embodiments are combined to provide reference. Figures 1 to 3 The technical effects achieved by the described embodiments are explained above. For details, please refer to the above description. Figures 1 to 3 The specific details of the description will not be repeated here.
[0100] The following is for reference Figure 5 A point cloud clustering method according to an embodiment of the present disclosure is described. Figure 5 A flowchart illustrating a point cloud clustering method according to an embodiment of the present disclosure is shown. Figure 5 As shown, the point cloud clustering method includes steps S501, S502, S503, and S504.
[0101] In step S501, the point cloud is projected onto the parameter space through a preset transformation method according to the preset parameter model to obtain the voxels corresponding to the point cloud in the parameter space.
[0102] In step S502, a specific voxel is selected from the voxels in the parameter space according to a preset voxel selection rule.
[0103] In step S503, based on the correspondence between the point cloud and the voxels in the parameter space, points corresponding to the specific voxels are selected from the point cloud as seed points for point cloud clustering.
[0104] In step S504, point cloud clustering is performed using the seed points.
[0105] According to the technical solution provided in this disclosure, a point cloud is projected onto a parameter space through a preset transformation method based on a preset parameter model to obtain the voxels corresponding to the point cloud in the parameter space; a specific voxel is selected from the voxels in the parameter space according to a preset voxel selection rule; based on the correspondence between the point cloud and the voxels in the parameter space, a point corresponding to the specific voxel is selected from the point cloud as a seed point for point cloud clustering; point cloud clustering is performed using the seed point, avoiding the problem of poor clustering effect caused by randomly selecting seed points. Moreover, it is not necessary to calculate the normal vector, which avoids the calculation of the feature matrix and the neighborhood search. Therefore, the computational complexity of O(nlogn) can be reduced to O(n), significantly reducing the amount of computation. At the same time, it avoids the problem of normal vector calculation being affected by noise.
[0106] Furthermore, the technical solution provided in this disclosure combines 3D Hough transform with a growing clustering algorithm, helping the growing algorithm to better select seed points and reducing the computational complexity from O(nlogn) to O(n), significantly reducing the computational load. In terms of performance, it avoids the problem of noise affecting normal vector calculation.
[0107] Those skilled in the art will understand that, with reference to Figure 5 The described technical solution can be compared with the reference. Figures 1 to 4 The described embodiments are combined to provide reference. Figures 1 to 4 The technical effects achieved by the described embodiments are explained above. For details, please refer to the above description. Figures 1 to 4 The specific details of the description will not be repeated here.
[0108] The following is for reference Figure 6 A sensor data processing method according to an embodiment of the present disclosure is described. Figure 6 A flowchart illustrating a sensor data processing method according to an embodiment of the present disclosure is shown. Figure 6 As shown, the sensor data processing method includes steps S601, S602, S603, and S604.
[0109] In step S601, the point cloud acquired by the sensor is projected onto the parameter space through a preset transformation method according to the preset parameter model to obtain the voxel corresponding to the point cloud in the parameter space.
[0110] In step S602, a specific voxel is selected from the voxels in the parameter space according to a preset voxel selection rule.
[0111] In step S603, based on the correspondence between the point cloud and the voxels in the parameter space, points corresponding to the specific voxels are selected from the point cloud as seed points for point cloud clustering.
[0112] In step S604, point cloud clustering is performed using the seed points.
[0113] In one embodiment of this disclosure, a sensor refers to a sensor capable of acquiring point cloud data, such as LiDAR, a sensor commonly used in autonomous driving and high-precision map acquisition. However, this disclosure is not limited to this, and any sensor capable of acquiring point cloud data can be used as a sensor in the embodiments of this disclosure.
[0114] According to the technical solution provided in this disclosure, point clouds acquired by sensors are projected onto a parameter space through a preset transformation method based on a preset parameter model to obtain voxels corresponding to the point clouds in the parameter space; specific voxels are selected from the voxels in the parameter space according to a preset voxel selection rule; based on the correspondence between the point clouds and the voxels in the parameter space, points corresponding to the specific voxels are selected from the point clouds as seed points for point cloud clustering; point cloud clustering is performed using the seed points, avoiding the problem of poor clustering results caused by randomly selecting seed points. Moreover, it is not necessary to calculate normal vectors, thus avoiding the calculation of feature matrices and neighborhood search. Therefore, the computational complexity can be reduced from O(nlogn) to O(n), significantly reducing the amount of computation. At the same time, it avoids the problem of normal vector calculation being affected by noise.
[0115] Furthermore, the technical solution provided in this disclosure combines 3D Hough transform with a growing clustering algorithm, helping the growing algorithm to better select seed points and reducing the computational complexity from O(nlogn) to O(n), significantly reducing the computational load when processing sensor data. In terms of performance, it avoids the problem of noise affecting normal vector calculation.
[0116] Those skilled in the art will understand that, with reference to Figure 6 The described technical solution can be compared with the reference. Figures 1 to 5 The described embodiments are combined to provide reference. Figures 1 to 5 The technical effects achieved by the described embodiments are explained above. For details, please refer to the above description. Figures 1 to 5 The specific details of the description will not be repeated here.
[0117] The following is for reference Figure 7 A sensor data processing apparatus according to an embodiment of the present disclosure is described. Figure 7 A block diagram of a sensor data processing apparatus 700 according to an embodiment of the present disclosure is shown.
[0118] like Figure 7 As shown, the sensor data processing device 700 includes:
[0119] The transformation module 701 is configured to project the point cloud acquired by the sensor into the parameter space through a preset transformation method according to the preset parameter model to obtain the voxel corresponding to the point cloud in the parameter space.
[0120] The voxel selection module 702 is configured to select a specific voxel from the voxels in the parameter space according to a preset voxel selection rule;
[0121] The seed point selection module 703 is configured to select points corresponding to a specific voxel from the point cloud as seed points for point cloud clustering based on the correspondence between the point cloud and the voxels in the parameter space.
[0122] Clustering module 704 is configured to perform point cloud clustering using the seed points.
[0123] According to the technical solution provided in this disclosure, a transformation module is configured to project the point cloud acquired by the sensor onto a parameter space using a preset transformation method based on a preset parameter model to obtain voxels corresponding to the point cloud in the parameter space; a voxel selection module is configured to select specific voxels from the voxels in the parameter space according to preset voxel selection rules; a seed point selection module is configured to select points corresponding to the specific voxels from the point cloud as seed points for point cloud clustering based on the correspondence between the point cloud and the voxels in the parameter space; and a clustering module is configured to perform point cloud clustering using the seed points, thus avoiding the problem of poor clustering results caused by randomly selecting seed points. Furthermore, it avoids calculating normal vectors, thereby avoiding the calculation of the feature matrix and neighborhood search. Therefore, the computational complexity can be reduced from O(nlogn) to O(n), significantly reducing the computational load. Simultaneously, it avoids the problem of noise affecting normal vector calculation.
[0124] Those skilled in the art will understand that, with reference to Figure 7 The described technical solution can be compared with the reference. Figures 1 to 6 The described embodiments are combined to provide reference. Figures 1 to 6 The technical effects achieved by the described embodiments are explained above. For details, please refer to the above description. Figures 1 to 6 The specific details of the description will not be repeated here.
[0125] Figure 8 A structural block diagram of an electronic device according to an embodiment of the present disclosure is shown.
[0126] This disclosure also provides an electronic device, such as... Figure 8As shown, it includes at least one processor 801; and a memory 802 communicatively connected to at least one processor 801; wherein the memory 802 stores instructions executable by at least one processor 801, the instructions being executed by at least one processor 801 to perform the following steps:
[0127] The point cloud is projected onto the parameter space according to the preset parameter model through a preset transformation method to obtain the voxel corresponding to the point cloud in the parameter space.
[0128] Specific voxels are selected from the voxels in the parameter space according to preset voxel selection rules;
[0129] Based on the correspondence between the point cloud and the voxels in the parameter space, points corresponding to the specific voxels are selected from the point cloud as seed points for point cloud clustering.
[0130] In one embodiment of this disclosure, the preset transformation method is the Hough transform.
[0131] In one embodiment of this disclosure, the step of projecting a point cloud onto a parameter space using a preset transformation method according to a preset parameter model to obtain the voxels corresponding to the point cloud in the parameter space includes:
[0132] The point cloud is projected into the parameter space through the Hough transform according to the preset planar model to obtain the voxels corresponding to the point cloud in the Hough parameter space.
[0133] In one embodiment of this disclosure, selecting a specific voxel from the voxels in the parameter space according to a preset voxel selection rule includes:
[0134] Determine the density of the voxels corresponding to the unclustered points in the point cloud in the parameter space;
[0135] Select the voxel with the highest density from the voxels as the specific voxel corresponding to the unclustered point in the point cloud.
[0136] This disclosure also provides an electronic device, such as... Figure 8 As shown, it includes at least one processor 801; and a memory 802 communicatively connected to at least one processor 801; wherein the memory 802 stores instructions executable by at least one processor 801, the instructions being executed by at least one processor 801 to perform the following steps:
[0137] The point cloud is projected onto the parameter space according to the preset parameter model through a preset transformation method to obtain the voxel corresponding to the point cloud in the parameter space.
[0138] Specific voxels are selected from the voxels in the parameter space according to preset voxel selection rules;
[0139] Based on the correspondence between the point cloud and the voxels in the parameter space, points corresponding to the specific voxels are selected from the point cloud as seed points for point cloud clustering.
[0140] Point cloud clustering is performed using the seed points.
[0141] Figure 9 This is a schematic diagram of the structure of a computer system suitable for implementing the methods according to the embodiments of this disclosure. For example... Figure 9 As shown, the computer system 900 includes a processing unit 901, which can execute various processes in the embodiments shown in the figures above, based on a program stored in a read-only memory (ROM) 902 or a program loaded from a storage section 908 into a random access memory (RAM) 903. The RAM 903 also stores various programs and data required for the operation of the system 900. The CPU 901, ROM 902, and RAM 903 are interconnected via a bus 904. An input / output (I / O) interface 905 is also connected to the bus 904.
[0142] The following components are connected to I / O interface 905: an input section 906 including a keyboard, mouse, etc.; an output section 907 including a cathode ray tube (CRT), liquid crystal display (LCD), etc., and a speaker, etc.; a storage section 908 including a hard disk, etc.; and a communication section 909 including a network interface card such as a LAN card, modem, etc. The communication section 909 performs communication processing via a network such as the Internet. A drive 910 is also connected to I / O interface 905 as needed. A removable medium 911, such as a disk, optical disk, magneto-optical disk, semiconductor memory, etc., is installed on drive 910 as needed so that computer programs read from it can be installed into storage section 908 as needed. The processing unit 901 can be implemented as a CPU, GPU, TPU, FPGA, NPU, etc.
[0143] In particular, according to embodiments of this disclosure, the methods described above with reference to the accompanying drawings can be implemented as computer software programs. For example, embodiments of this disclosure include a computer program product comprising a computer program tangibly embodied on a readable medium thereof, the computer program containing program code for performing the methods in the drawings. In such embodiments, the computer program can be downloaded and installed from a network via a communication section 909, and / or installed from a removable medium 911. For example, embodiments of this disclosure include a readable storage medium storing computer instructions that, when executed by a processor, implement program code for performing the methods in the drawings.
[0144] The flowcharts and block diagrams in the accompanying drawings illustrate the architecture, functionality, and operation of possible implementations of systems, methods, and computer program products according to various embodiments of this disclosure. In this regard, each block in a flowchart or block diagram may represent a module, segment, or portion of code containing one or more executable instructions for implementing a specified logical function. It should also be noted that in some alternative implementations, the functions indicated in the blocks may occur in a different order than those indicated in the drawings. For example, two consecutively indicated blocks may actually be executed substantially in parallel, and they may sometimes be executed in reverse order, depending on the functions involved. It should also be noted that each block in the block diagrams and / or flowcharts, and combinations of blocks in the block diagrams and / or flowcharts, can be implemented using a dedicated hardware-based system that performs the specified function or operation, or using a combination of dedicated hardware and computer instructions.
[0145] The units or modules described in the embodiments of this disclosure can be implemented in software or hardware. The described units or modules can also be located in a processor, and the names of these units or modules do not necessarily constitute a limitation on the unit or module itself.
[0146] In another aspect, this disclosure also provides a computer-readable storage medium, which may be the computer-readable storage medium included in the nodes described in the above embodiments; or it may be a standalone computer-readable storage medium not assembled into a device. The computer-readable storage medium stores one or more programs that are used by one or more processors to perform the methods described in this disclosure.
[0147] The above description is merely a preferred embodiment of this disclosure and an explanation of the technical principles employed. Those skilled in the art should understand that the scope of the invention involved in this disclosure is not limited to technical solutions formed by specific combinations of the above-described technical features, but should also cover other technical solutions formed by arbitrary combinations of the above-described technical features or their equivalents without departing from the inventive concept. For example, technical solutions formed by substituting the above-described features with (but not limited to) technical features disclosed in this disclosure that have similar functions.
Claims
1. A method for selecting seed points in a point cloud, comprising: The point cloud is projected onto the parameter space according to the preset parameter model through a preset transformation method to obtain the voxels corresponding to the point cloud in the parameter space, wherein the preset transformation method is the Hough transform. Specific voxels are selected from the voxels in the parameter space according to preset voxel selection rules; Based on the correspondence between the point cloud and the voxels in the parameter space, points corresponding to the specific voxels are selected from the point cloud as seed points for point cloud clustering.
2. The method of claim 1, wherein, The step of projecting the point cloud onto the parameter space using a preset transformation method according to a preset parameter model to obtain the voxels corresponding to the point cloud in the parameter space includes: The point cloud is projected into the parameter space through the Hough transform according to the preset planar model to obtain the voxels corresponding to the point cloud in the Hough parameter space.
3. The method of claim 1 or 2, wherein, The step of selecting a specific voxel from the voxels in the parameter space according to a preset voxel selection rule includes: Determine the density of the voxels corresponding to the unclustered points in the point cloud in the parameter space; Select the voxel with the highest density from the voxels as the specific voxel corresponding to the unclustered point in the point cloud.
4. A point cloud seed point selection device, comprising: The transformation module is configured to project the point cloud onto the parameter space through a preset transformation method according to the preset parameter model to obtain the voxels corresponding to the point cloud in the parameter space, wherein the preset transformation method is the Hough transform. The voxel selection module is configured to select a specific voxel from the voxels in the parameter space according to a preset voxel selection rule; The seed point selection module is configured to select points corresponding to a specific voxel from the point cloud as seed points for point cloud clustering based on the correspondence between the point cloud and the voxels in the parameter space.
5. The apparatus of claim 4, wherein, The voxel selection module includes: The determination submodule is configured to determine the density of voxels corresponding to the unclustered points in the point cloud in the parameter space; The voxel selection submodule is configured to select the voxel with the highest density from the voxels as the specific voxel corresponding to the unclustered point in the point cloud.
6. A point cloud clustering method, comprising: The point cloud is projected onto the parameter space according to the preset parameter model through a preset transformation method to obtain the voxels corresponding to the point cloud in the parameter space, wherein the preset transformation method is the Hough transform. Specific voxels are selected from the voxels in the parameter space according to preset voxel selection rules; Based on the correspondence between the point cloud and the voxels in the parameter space, points corresponding to the specific voxels are selected from the point cloud as seed points for point cloud clustering. Point cloud clustering is performed using the seed points.
7. An electronic device comprising a memory and a processor; wherein, The memory is used to store one or more computer instructions, wherein the one or more computer instructions are executed by the processor to implement the method as described in any one of claims 1-3, 6.
8. A readable storage medium having stored thereon computer instructions that, when executed by a processor, implement the method as claimed in any one of claims 1-3 and 6.
9. A computer program product comprising computer instructions which, when executed by a processor, implement the method of any of claims 1-3, 6.