Multi-scale point cloud feature extraction method and device for dividing neighborhoods based on feature points

Abstract

The invention is suitable for the technical field of point cloud data processing, and provides a multi-scale point cloud feature extraction method and device based on neighborhood division of feature points. In the embodiment of the invention, for the target point cloud data needing feature extraction, the density change value and the density change gradient of the point cloud data are firstly determined, so that the target feature point in the point cloud data is determined. Determining a neighborhood feature point set of each target feature point according to different radiuses, sampling feature points in the neighborhood feature point set, calculating a feature vector of each sampled feature point, and combining the feature points into a target feature vector; and finally, determining the multi-scale features of the target point cloud data through a plurality of target feature vectors determined according to different radiuses. The problems that an existing point cloud feature extraction method is poor in fine geometric feature capturing effect, high in calculation cost and low in calculation efficiency are solved.

Description

Technical Field

[0001] The present application relates to the technical field of point cloud data processing, and in particular to a multi-scale point cloud feature extraction method and device based on feature point neighborhood division. Background Art

[0002] Feature extraction of point cloud data is an important task in 3D data processing and is widely used in fields such as autonomous driving and robotics. Its main goal is to extract meaningful feature information from sparse and unordered 3D point cloud data. Methods that directly process point cloud data have received widespread attention in recent years. Representative technologies include PointNet and its derivatives. PointNet is the first method to directly process unordered point cloud data. It extracts features from each point separately and then integrates global features using symmetry functions (such as max-pooling). PointNet++ further improves this architecture and introduces a hierarchical sampling mechanism to capture local geometric features, thereby improving the model's ability to describe details. These methods do not need to convert point cloud data into grids or voxels, avoiding information loss and computational overhead caused by data conversion. In addition, there are other technologies such as DGCNN, which mine the relationship between local points in point cloud data by constructing a dynamic neighborhood graph. These methods that directly process point clouds have good flexibility and scalability and are suitable for processing complex 3D data.

[0003] Although methods that directly process point cloud data (such as PointNet, PointNet++, and DGCNN) have made significant progress in 3D point cloud processing, there are still some objective shortcomings. First, PointNet processes point clouds point by point, lacks the characterization of local structures, and cannot make full use of neighborhood information, so it is not effective in capturing subtle geometric features. Although PointNet++ has improved this through layered sampling and local feature extraction, its local area division still relies on Euclidean distance and is difficult to cope with scenes with complex topological structures. Secondly, although DGCNN introduces dynamic neighborhood composition, it requires a high computational cost due to the dynamic changes of the neighborhood, especially when processing large-scale point cloud data, its computational efficiency may be low. Although these technologies avoid the information loss caused by data conversion, there is still room for improvement in processing high-dimensional data and complex geometric structures. Summary of the invention

[0004] In view of this, the present application provides a multi-scale point cloud feature extraction method and device based on feature point neighborhood division to solve the problems of poor effect of existing point cloud feature extraction methods in capturing subtle geometric features, high computational cost and low computational efficiency.

[0005] The first aspect of the present application provides a multi-scale point cloud feature extraction method based on feature point neighborhood division, the method comprising: Acquire target point cloud data, determine neighborhood feature point sets of any feature point in the target point cloud data at different radii according to a preset radius set r, and determine feature point density of the neighborhood feature point set; Determine the density change value of the feature point according to the neighborhood feature point set, the radius set R and the corresponding feature point density and density gradient , where the density change value Characterizes the density change of the neighborhood feature point set generated by the feature point at different radii, and the density change gradient Characterize the density difference between the feature point and other feature points in the neighborhood feature point set corresponding to the feature point; According to the density change value and the density gradient Determine the target feature point set , for the target feature point set Any target feature point within the preset initial radius and determining a target neighborhood feature point set of the target feature point based on a preset feature point quantity interval; Determine a number of sampling feature points from each target neighborhood feature point set, calculate the feature vector of each sampling feature point, merge the feature vectors into the target feature vector, and perform several adjustments to the feature point quantity interval according to the preset initial radius. and the adjusted feature point quantity interval to determine the target feature vector, and determine the multi-scale features of the target point cloud data according to the target feature vector and the adjusted target feature vector.

[0006] Optionally, the density change value of the feature point is determined according to the neighborhood feature point set, the radius set r and the corresponding feature point density. include: By formula Determine the feature points Density change value , wherein the The radius Lower feature point The feature point density of the corresponding neighborhood feature point set.

[0007] Optionally, the density change gradient is determined according to the neighborhood feature point set, the radius set r and the corresponding feature point density. include: By formula Determine feature points With the feature point The density difference of other feature points in the neighborhood feature point set, where is the neighborhood feature point set, is the number of feature points in the neighborhood feature point set, is the feature point density of the neighborhood feature point set corresponding to the feature point, is the feature point density of the neighborhood feature point set corresponding to other feature points.

[0008] Optionally, determining a target neighborhood feature point set of the target feature point according to a preset initial radius and a preset feature point quantity interval includes: For any target feature point, determine an initial neighborhood feature point set of the target feature point according to a preset initial radius, and compare the number of feature points in the initial neighborhood feature point set with a preset feature point number interval; When the number of feature points is less than the minimum value of the feature point number interval, increase the preset initial radius, and re-determine the number of feature points of the initial neighborhood feature point set of the target feature point according to the increased preset initial radius until the number of feature points is within the feature point number interval; When the number of feature points is greater than the maximum value of the feature point number interval, the preset initial radius is reduced, and the number of feature points of the initial neighborhood feature point set of the target feature point is re-determined according to the reduced preset initial radius until the number of feature points is within the feature point number interval.

[0009] Optionally, determining the multi-scale features of the target point cloud data according to the target feature vector and the adjusted target feature vector includes: By formula The target feature vectors determined by the number intervals of each feature point are cascaded to obtain the multi-scale feature, wherein the The initial radius is The target feature vector is determined when .

[0010] The second aspect of the present application provides a multi-scale point cloud feature extraction device based on feature point neighborhood division, the device comprising: A density determination unit, used to obtain target point cloud data, determine a neighborhood feature point set of any feature point in the target point cloud data at different radii according to a preset radius set r, and determine a feature point density of the neighborhood feature point set; A density change parameter determination unit, configured to determine the density change value of the feature point according to the neighborhood feature point set, the radius set R and the corresponding feature point density. and density gradient , where the density change value Characterizes the density change of the neighborhood feature point set generated by the feature point at different radii, and the density change gradient Characterize the density difference between the feature point and other feature points in the neighborhood feature point set corresponding to the feature point; The target neighborhood feature point set determination unit is used to determine the target neighborhood feature point set according to the density change value. and the density gradient Determine the target feature point set , for the target feature point set Any target feature point within the preset initial radius and determining a target neighborhood feature point set of the target feature point based on a preset feature point quantity interval; A multi-scale feature determination unit is used to determine a number of sampling feature points from each target neighborhood feature point set, calculate the feature vector of each sampling feature point, merge the feature vectors into a target feature vector, and perform several adjustments to the feature point quantity interval according to the preset initial radius. and the adjusted feature point quantity interval to determine the target feature vector, and determine the multi-scale features of the target point cloud data according to the target feature vector and the adjusted target feature vector.

[0011] Optionally, the density change parameter determination unit determines the density change value of the feature point according to the neighborhood feature point set, the radius set r and the corresponding feature point density. include: By formula Determine the feature points Density change value , wherein the The radius Lower feature point The feature point density of the corresponding neighborhood feature point set.

[0012] Optionally, the density change parameter determination unit determines the density change gradient according to the neighborhood feature point set, the radius set r and the corresponding feature point density. include: By formula Determine feature points With the feature point The density difference of other feature points in the neighborhood feature point set, where is the neighborhood feature point set, is the number of feature points in the neighborhood feature point set, is the feature point density of the neighborhood feature point set corresponding to the feature point, is the feature point density of the neighborhood feature point set corresponding to other feature points.

[0013] Optionally, the target neighborhood feature point set determining unit determines the target neighborhood feature point set of the target feature point according to a preset initial radius and a preset feature point quantity interval, including: For any target feature point, determine an initial neighborhood feature point set of the target feature point according to a preset initial radius, and compare the number of feature points in the initial neighborhood feature point set with a preset feature point number interval; When the number of feature points is less than the minimum value of the feature point number interval, increase the preset initial radius, and re-determine the number of feature points of the initial neighborhood feature point set of the target feature point according to the increased preset initial radius until the number of feature points is within the feature point number interval; When the number of feature points is greater than the maximum value of the feature point number interval, the preset initial radius is reduced, and the number of feature points of the initial neighborhood feature point set of the target feature point is re-determined according to the reduced preset initial radius until the number of feature points is within the feature point number interval.

[0014] Optionally, the determining of the multi-scale features of the target point cloud data according to the target feature vector and the adjusted target feature vector in the multi-scale feature determining unit includes: By formula The target feature vectors determined by the number intervals of each feature point are cascaded to obtain the multi-scale feature, wherein the The initial radius is The target feature vector is determined when .

[0015] In the embodiment provided in the present application, for the target point cloud data that needs to be feature extracted, the density change value and density change gradient of the point cloud data are first determined to determine the target feature points in the point cloud data. Then, the neighborhood feature point set of each target feature point is determined according to different radii, and then the feature points in the neighborhood feature point set are sampled and the feature vector of each sampled feature point is calculated, which is then merged into a target feature vector, and finally the multi-scale features of the target point cloud data are determined by a number of target feature vectors determined by different radii.

[0016] The advantages of this application are: 1) Adapting to the local topological structure of the point cloud: This application can better adapt to the complex topological structure of the point cloud data by first selecting feature points and then dividing the neighborhood, overcoming the defect of the traditional method of evenly dividing the neighborhood that ignores the geometric characteristics of the point cloud.

[0017] 2) Dynamically adjust the neighborhood radius: Dynamically adjust the neighborhood radius according to the local density of feature points to ensure that there are a reasonable number of feature points in the neighborhood of each feature point, thereby improving the representation of neighborhood information. This helps avoid errors in sparse or overcrowded areas and enhances the extraction of geometric structure information.

[0018] 3) Multi-scale feature extraction: Multi-scale density calculation and feature extraction are used to capture multi-level geometric information in point cloud data through neighborhoods of different radii, making the extracted features more complete and enhancing the expressiveness of point cloud features. BRIEF DESCRIPTION OF THE DRAWINGS

[0019] Figure 1 A flow chart of the method provided in the embodiment of the present application; Figure 2 A schematic diagram of neighborhood division provided in an embodiment of the present application; Figure 3 A schematic diagram of feature extraction provided in an embodiment of the present application; Figure 4 A structural diagram of a device provided in an embodiment of the present application; Figure 5 A schematic diagram of the internal structure of a computer device provided in an embodiment of the present application. DETAILED DESCRIPTION

[0020] Exemplary embodiments will be described in detail herein, examples of which are shown in the accompanying drawings. When the following description refers to the drawings, the same numbers in different drawings represent the same or similar elements unless otherwise indicated. The implementations described in the following exemplary embodiments do not represent all implementations consistent with the present application. Instead, they are merely examples of devices and methods consistent with some aspects of the present application as detailed in the appended claims.

[0021] The terms used in this application are for the purpose of describing specific embodiments only and are not intended to limit this application. The singular forms of "a", "said" and "the" used in this application and the appended claims are also intended to include plural forms unless the context clearly indicates other meanings. It should also be understood that the term "and / or" used herein refers to and includes any or all possible combinations of one or more associated listed items.

[0022] It should be understood that although the terms first, second, third, etc. may be used in the present application to describe various information, these information should not be limited to these terms. These terms are only used to distinguish the same type of information from each other. For example, without departing from the scope of the present application, the first information may also be referred to as the second information, and similarly, the second information may also be referred to as the first information. Depending on the context, the word "if" as used herein may be interpreted as "at the time of" or "when" or "in response to determining".

[0023] The present application provides a multi-scale point cloud feature extraction method and device based on feature point neighborhood division to solve the problems of poor effect of existing point cloud feature extraction methods in capturing subtle geometric features, high computational cost and low computational efficiency.

[0024] The technical solution of the present application is described in detail with specific embodiments below. The following specific embodiments can be combined with each other, and the same or similar concepts or processes may not be described in detail in some embodiments.

[0025] like Figure 1 As shown, it is a flowchart of a multi-scale point cloud feature extraction method based on feature point neighborhood division provided by the present application, and the steps of the process include: S101, acquiring target point cloud data, determining neighborhood feature point sets of any feature point in the target point cloud data at different radii according to a preset radius set r, and determining feature point density of the neighborhood feature point set.

[0026] In this embodiment, for each feature point in the target point cloud data , from the radius set r={ , ,… }Select a value as the radius of the neighborhood, find all the feature points within the neighborhood, and thus determine the neighborhood feature point set, and pass The density of the neighborhood feature point set is expressed by the formula Determine, among which, is the number of feature points in the neighborhood feature point set, is the volume of the neighborhood sphere, which can be obtained by To calculate. Feature point density Indicate point The density of feature points in its neighborhood.

[0027] Traverse each radius in the radius set r and each feature point in the target point cloud data, for each point , can be at multiple different radii The density of feature points of the neighborhood feature point set is calculated as follows: By changing the density at different scales, the changes in the local geometric structure of feature points can be captured.

[0028] S102, determining the density change value of the feature point according to the neighborhood feature point set, the radius set r and the corresponding feature point density and density gradient , where the density change value Characterizes the density change of the neighborhood feature point set generated by the feature point at different radii, and the density change gradient Characterizes the density difference between the feature point and other feature points in the neighborhood feature point set corresponding to the feature point.

[0029] In this embodiment, the density change value The formula can be OK, in this formula At different radii Lower feature point The density of feature points in the neighborhood feature point set. Density change value It can reflect the density difference of point clouds at different scales. A larger density change value means that the point At prominent locations in the local geometry, such as edges or inflection points.

[0030] For density gradient By comparing the density difference between a feature point and other feature points in its neighborhood, the density change gradient of the feature point is obtained. The specific calculation formula is as follows: , in this formula is the neighborhood feature point set, is the number of feature points in the neighborhood feature point set, is the density of feature points of the neighborhood feature point set, The density of feature points in the neighborhood feature point set corresponding to other feature points in the neighborhood feature point set. Determine the density of feature points and The method is the same as the above method, so it will not be repeated here. The density gradient can be used as a feature point The direction of density change in its neighborhood helps determine the location of significant feature points.

[0031] S103, according to the density change value and the density gradient Determine the target feature point set , for the target feature point set Any target feature point within the preset initial radius And a preset feature point quantity interval is used to determine a target neighborhood feature point set of the target feature point.

[0032] In this embodiment, the density change value can be and density gradient Set a threshold value respectively, such as Corresponding density change value , Corresponding density gradient When the feature point , it means that the geometric structure of the area where the feature point is located is complex, for example, the curvature changes greatly. Then the feature point is defined as the target feature point , add the target feature point set . Or when the feature point When , it means that the density change direction of the feature point is more obvious, indicating that the feature point may be located at the edge or inflection point of the local geometric structure. Then the feature point is defined as the target feature point , add the target feature point set .

[0033] Determine the target feature point set Then, for each target feature point in the set , through the method in step S101, determine that each target feature point is within a preset initial radius The initial target neighborhood feature point set is obtained, and then the number of feature points in the neighborhood feature point set is compared with the preset feature point number interval; When the number of feature points is less than the minimum value of the feature point number interval, increase the preset initial radius , and according to the enlarged preset initial radius Re-determining the number of feature points of the initial neighborhood feature point set of the target feature point until the number of feature points is within the feature point number interval; When the number of feature points is greater than the maximum value of the feature point number interval, reduce the preset initial radius , and according to the reduced preset initial radius Re-determine the number of feature points of the initial neighborhood feature point set of the target feature point until the number of feature points is within the feature point number interval. The final neighborhood division diagram is as follows: Figure 2 shown.

[0034] Finally, the initial neighborhood feature point set whose number of feature points is within the quantity interval is determined as the target neighborhood feature point set, which is recorded as .

[0035] S104, determining a number of sampling feature points from each target neighborhood feature point set, calculating a feature vector of each sampling feature point, merging the feature vectors into a target feature vector, and performing a number of adjustments to the feature point quantity interval according to the preset initial radius. and the adjusted feature point quantity interval to determine the target feature vector, and determine the multi-scale features of the target point cloud data according to the target feature vector and the adjusted target feature vector.

[0036] In this embodiment, the target neighborhood feature point set generated for each target feature point can be recorded as , the PointNet method can be used to extract features from each target neighborhood feature point set. The specific method is as follows: 1. Use the FPS (Farthest Point Sampling) algorithm to determine the sampling feature points from a set of several target neighborhood feature points. The sampling feature points selected by this method can ensure that the sampling points in each area are evenly distributed, thereby reducing the computational complexity.

[0037] 2. Input each sampled feature point into the PointNet network, and apply the multi-layer perceptron (MLP) point by point to calculate the feature vector of each point: .in is the coordinate information of the i-th sampling point, is the local feature vector of the point. Then the point features in the neighborhood of each feature point are merged into a fixed-length target feature vector through global maximum pooling: .

[0038] After determining the target feature vector, in order to fully capture the multi-scale information of the point cloud, we first change the number of feature points in different intervals, and then repeat the steps of grouping, sampling, and PointNet feature extraction to extract features at multiple scales. This generates target feature vectors with different initial radii. , and then concatenate each target feature vector to finally obtain the multi-scale features of the target point cloud data, expressed as , the specific feature extraction diagram is as follows Figure 3 shown.

[0039] This embodiment dynamically adjusts the neighborhood radius according to the local density of feature points to ensure that there are a reasonable number of feature points in the neighborhood of each feature point, thereby improving the representation capability of neighborhood information. This helps avoid errors in sparse or overcrowded areas and enhances the extraction effect of geometric structure information.

[0040] Finally, the geometric structure of point clouds at different scales is effectively captured through multi-scale feature fusion method, which enhances the feature expression ability.

[0041] So far, completed Figure 1 The process shown.

[0042] In the embodiment of the present application, for the target point cloud data that needs to be feature extracted, the density change value and density change gradient of the point cloud data are first determined to determine the target feature points in the point cloud data. Then, the neighborhood feature point set of each target feature point is determined according to different radii, and then the feature points in the neighborhood feature point set are sampled and the feature vector of each sampled feature point is calculated, which is then merged into a target feature vector. Finally, the multi-scale features of the target point cloud data are determined by several target feature vectors determined by different radii. This solves the problem that the existing point cloud feature extraction methods are not effective in capturing subtle geometric features, have high computational costs, and have low computational efficiency.

[0043] The present application also provides a multi-scale point cloud feature extraction device based on feature point neighborhood division, such as Figure 4 As shown, the device comprises: Density determination unit 401, used to obtain target point cloud data, according to the preset radius set r={ , ,… Determine a neighborhood feature point set of any feature point in the target point cloud data at different radii, and determine the feature point density of the neighborhood feature point set; The density change parameter determination unit 402 is used to determine the density change value of the feature point according to the neighborhood feature point set, the radius set R and the corresponding feature point density. and density gradient , where the density change value Characterizes the density change of the neighborhood feature point set generated by the feature point at different radii, and the density change gradient Characterize the density difference between the feature point and other feature points in the neighborhood feature point set corresponding to the feature point; The target neighborhood feature point set determination unit 403 is configured to determine the target neighborhood feature point set according to the density change value. and the density gradient Determine the target feature point set , for the target feature point set Any target feature point within the preset initial radius and determining a target neighborhood feature point set of the target feature point based on a preset feature point quantity interval; The multi-scale feature determination unit 404 is used to determine a number of sampling feature points from each target neighborhood feature point set, calculate the feature vector of each sampling feature point, merge the feature vector into a target feature vector, and perform several adjustments to the feature point quantity interval according to the preset initial radius. and the adjusted feature point quantity interval to determine the target feature vector, and determine the multi-scale features of the target point cloud data according to the target feature vector and the adjusted target feature vector.

[0044] In another embodiment, the density change parameter determination unit determines the density change value of the feature point according to the neighborhood feature point set, the radius set R and the corresponding feature point density. include: By formula Determine the feature points Density change value , wherein the The radius Lower feature point The feature point density of the corresponding neighborhood feature point set.

[0045] In another embodiment, the density change parameter determination unit determines the density change gradient according to the neighborhood feature point set, the radius set R and the corresponding feature point density. include: By formula Determine feature points With the feature point The density difference of other feature points in the neighborhood feature point set, where is the neighborhood feature point set, is the number of feature points in the neighborhood feature point set, is the feature point density of the neighborhood feature point set corresponding to the feature point, is the feature point density of the neighborhood feature point set corresponding to other feature points.

[0046] In another embodiment, the target neighborhood feature point set determination unit determines the target neighborhood feature point set of the target feature point according to the preset initial radius and the preset feature point quantity interval, including: For any target feature point, determine an initial neighborhood feature point set of the target feature point according to a preset initial radius, and compare the number of feature points in the initial neighborhood feature point set with a preset feature point number interval; When the number of feature points is less than the minimum value of the feature point number interval, increase the preset initial radius, and re-determine the number of feature points of the initial neighborhood feature point set of the target feature point according to the increased preset initial radius until the number of feature points is within the feature point number interval; When the number of feature points is greater than the maximum value of the feature point number interval, the preset initial radius is reduced, and the number of feature points of the initial neighborhood feature point set of the target feature point is re-determined according to the reduced preset initial radius until the number of feature points is within the feature point number interval.

[0047] In another embodiment, determining the multi-scale features of the target point cloud data according to the target feature vector and the adjusted target feature vector in the multi-scale feature determination unit includes: By formula The target feature vectors determined by the number intervals of each feature point are cascaded to obtain the multi-scale feature, wherein the The initial radius is The target feature vector is determined when .

[0048] The above-mentioned embodiment of the present invention provides a multi-scale point cloud feature extraction method based on feature point neighborhood division, and based on this method, provides a multi-scale point cloud feature extraction device based on feature point neighborhood division. Through the above-mentioned method and device, the problems of poor effect of existing point cloud feature extraction methods in capturing subtle geometric features, high computational cost and low computational efficiency can be solved.

[0049] This embodiment also discloses a computer device, such as Figure 5 As shown, the computer device includes a processor and a memory, wherein the memory stores at least one instruction, and the at least one instruction is loaded and executed by the processor to implement any of the above-mentioned multi-scale point cloud feature extraction methods based on feature point neighborhood division.

[0050] The above description is only a preferred embodiment of the present application and is not intended to limit the present application. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present application shall be included in the scope of protection of the present application.

Claims

1. A multi-scale point cloud feature extraction method based on feature point neighborhood division, characterized in that: The method comprises: Acquire target point cloud data, determine neighborhood feature point sets of any feature point in the target point cloud data at different radii according to a preset radius set r, and determine feature point density of the neighborhood feature point set; Determine the density change value of the feature point according to the neighborhood feature point set, the radius set r and the corresponding feature point density and density gradient , wherein the density change value Characterizes the density change of the neighborhood feature point set generated by the feature point at different radii, and the density change gradient Characterize the density difference between the feature point and other feature points in the neighborhood feature point set corresponding to the feature point; According to the density change value and the density gradient Determine the target feature point set , for the target feature point set Any target feature point within the preset initial radius and determining a target neighborhood feature point set of the target feature point based on a preset feature point quantity interval; Determine a number of sampling feature points from each target neighborhood feature point set, calculate the feature vector of each sampling feature point, merge the feature vectors into the target feature vector, and perform several adjustments to the feature point quantity interval according to the preset initial radius. and the adjusted feature point quantity interval to determine the target feature vector, and determine the multi-scale features of the target point cloud data according to the target feature vector and the adjusted target feature vector.

2. The method according to claim 1, characterized in that The density change value of the feature point is determined according to the neighborhood feature point set, the radius set r and the corresponding feature point density. include: By formula Determine the feature points Density change value , wherein the The radius Lower feature point The feature point density of the corresponding neighborhood feature point set.

3. The method according to claim 1, characterized in that The density change gradient is determined according to the neighborhood feature point set, the radius set r and the corresponding feature point density. include: By formula Determine feature points With the feature point The density difference of other feature points in the neighborhood feature point set, where is the neighborhood feature point set, is the number of feature points in the neighborhood feature point set, is the feature point density of the neighborhood feature point set corresponding to the feature point, is the feature point density of the neighborhood feature point set corresponding to other feature points.

4. The method according to claim 1, characterized in that: The step of determining a target neighborhood feature point set of the target feature point according to a preset initial radius and a preset feature point quantity interval comprises: For any target feature point, determine an initial neighborhood feature point set of the target feature point according to a preset initial radius, and compare the number of feature points in the initial neighborhood feature point set with a preset feature point number interval; When the number of feature points is less than the minimum value of the feature point number interval, increase the preset initial radius, and re-determine the number of feature points of the initial neighborhood feature point set of the target feature point according to the increased preset initial radius until the number of feature points is within the feature point number interval; When the number of feature points is greater than the maximum value of the feature point number interval, the preset initial radius is reduced, and the number of feature points of the initial neighborhood feature point set of the target feature point is re-determined according to the reduced preset initial radius until the number of feature points is within the feature point number interval.

5. The method according to claim 1, characterized in that: Determining the multi-scale features of the target point cloud data according to the target feature vector and the adjusted target feature vector includes: By formula The target feature vectors determined by the number intervals of each feature point are cascaded to obtain the multi-scale feature, wherein the The initial radius is The target feature vector is determined when .

6. A multi-scale point cloud feature extraction device based on feature point neighborhood division, characterized in that: The device comprises: A density determination unit, used to obtain target point cloud data, determine a neighborhood feature point set of any feature point in the target point cloud data at different radii according to a preset radius set r, and determine a feature point density of the neighborhood feature point set; A density change parameter determination unit, configured to determine the density change value of the feature point according to the neighborhood feature point set, the radius set R and the corresponding feature point density. and density gradient , wherein the density change value Characterizes the density change of the neighborhood feature point set generated by the feature point at different radii, and the density change gradient Characterize the density difference between the feature point and other feature points in the neighborhood feature point set corresponding to the feature point; The target neighborhood feature point set determination unit is used to determine the target neighborhood feature point set according to the density change value. and the density gradient Determine the target feature point set , for the target feature point set Any target feature point within the preset initial radius and determining a target neighborhood feature point set of the target feature point based on a preset feature point quantity interval; A multi-scale feature determination unit is used to determine a number of sampling feature points from each target neighborhood feature point set, calculate the feature vector of each sampling feature point, merge the feature vectors into a target feature vector, and perform several adjustments to the feature point quantity interval according to the preset initial radius. and the adjusted feature point quantity interval to determine the target feature vector, and determine the multi-scale features of the target point cloud data according to the target feature vector and the adjusted target feature vector.

7. The device according to claim 6, characterized in that The density change parameter determination unit determines the density change value of the feature point according to the neighborhood feature point set, the radius set R and the corresponding feature point density. include: By formula Determine the feature points Density change value , wherein the The radius Lower feature point The feature point density of the corresponding neighborhood feature point set.

8. The device according to claim 6, characterized in that The density change parameter determination unit determines the density change gradient according to the neighborhood feature point set, the radius set R and the corresponding feature point density. include: By formula Determine feature points With the feature point The density difference of other feature points in the neighborhood feature point set, where is the neighborhood feature point set, is the number of feature points in the neighborhood feature point set, is the feature point density of the neighborhood feature point set corresponding to the feature point, is the feature point density of the neighborhood feature point set corresponding to other feature points.

9. The device according to claim 6, characterized in that The target neighborhood feature point set determination unit determines the target neighborhood feature point set of the target feature point according to the preset initial radius and the preset feature point quantity interval, including: For any target feature point, determine an initial neighborhood feature point set of the target feature point according to a preset initial radius, and compare the number of feature points in the initial neighborhood feature point set with a preset feature point number interval; When the number of feature points is less than the minimum value of the feature point number interval, increase the preset initial radius, and re-determine the number of feature points of the initial neighborhood feature point set of the target feature point according to the increased preset initial radius until the number of feature points is within the feature point number interval; When the number of feature points is greater than the maximum value of the feature point number interval, the preset initial radius is reduced, and the number of feature points of the initial neighborhood feature point set of the target feature point is re-determined according to the reduced preset initial radius until the number of feature points is within the feature point number interval.

10. The device according to claim 6, characterized in that Determining the multi-scale features of the target point cloud data according to the target feature vector and the adjusted target feature vector in the multi-scale feature determination unit includes: By formula The target feature vectors determined by the number intervals of each feature point are cascaded to obtain the multi-scale feature, wherein the The initial radius is The target feature vector is determined when .

Images