Road scene inter-frame point cloud determines the nearest neighbor point method, storage medium and robot

Abstract

The application relates to a road scene interframe point cloud nearest neighbor point determination method, a storage medium and a robot, and comprises the following steps: S1: acquiring point cloud data P and Q of two frames before and after a laser radar at a time t, and performing directional segmentation on the two frames of point cloud P and Q; S2: removing a non-common interval in the two frames of point cloud; S3: performing nearest neighbor point matching on points in a corresponding interval of the common part; S4: selecting corresponding points in a case that an average Euclidean distance is minimum, and then the nearest neighbor point determination is successful. The method has the following advantages: by performing interval segmentation on the point cloud, the common part of the point cloud is retained, the non-common part is removed, the calculation content is simplified, invalid calculation is omitted, and the probability of occurrence of error corresponding points is reduced; the running efficiency can be improved without reducing the subsequent matching accuracy, and the real-time performance of the point cloud matching algorithm is improved.

Description

Technical Field

[0001] This invention relates to the field of robot navigation technology, and in particular to a method for determining nearest neighbor points in inter-frame point clouds of a road scene, a storage medium, and a robot. Background Technology

[0002] Because the robot moves at a high speed in road scenes, it generates a lot of point cloud data, and there are many non-common point clouds that are not needed in the stage of determining the nearest neighbor.

[0003] Determining the nearest neighbor is the first step in point cloud registration, establishing a one-to-one correspondence between points in the common area of ​​two point clouds. Currently, the nearest neighbor between the source and target point clouds is determined by calculating and comparing the average Euclidean distance between points in the common area. Given a source point cloud P containing n points and a target point cloud Q containing m points, m×n possible point cloud correspondences are generated. The average Euclidean distance for all cases is calculated, resulting in m×n possible outcomes. The case with the smallest average Euclidean distance is selected as the correspondence between the points in the two frames.

[0004] The disadvantages of current technologies:

[0005] 1. Existing technologies require calculating m×n possible outcomes to determine the corresponding point, which is only suitable for cases with a small number of point clouds. When the number of point clouds is large, the number of possible outcomes increases dramatically, resulting in excessive computation and significant time consumption. For example, if two adjacent frames each contain 4 point clouds, there are 4 possible outcomes for matching the nearest neighbor. 4 = 16 kinds.

[0006] 2. When the outdoor robot moves at a high speed, there are many non-common parts that do not participate in determining the corresponding points between two frames of point cloud, resulting in a lot of invalid calculations in subsequent steps.

[0007] 3. When dealing with large and complex point cloud data, a certain number of incorrect corresponding points may be generated during the calculation process.

[0008] The preceding description is intended to provide general background information and does not necessarily constitute prior art. Summary of the Invention

[0009] The purpose of this invention is to provide a method for determining the nearest neighbor point in a road scene point cloud between frames, a storage medium, and a robot. This method for determining the nearest neighbor point in a road scene point cloud between frames can improve operating efficiency and enhance the real-time performance of the point cloud matching algorithm without reducing the accuracy of subsequent matching.

[0010] This invention provides a method for determining nearest neighbor points in point clouds of a road scene across frames, comprising the following steps:

[0011] S1: Obtain point cloud data P and Q from the lidar before and after time t, and perform directional segmentation on the point clouds of frames P and Q.

[0012] S2: Remove non-common regions from the two point clouds;

[0013] S3: Perform nearest neighbor matching on points within the corresponding interval of the common part;

[0014] S4: Select the point corresponding to the case with the minimum average Euclidean distance, and the nearest neighboring point will be successfully determined.

[0015] Further, step S1 includes:

[0016] S11: Name the two point cloud frames acquired by the lidar as set P and set Q;

[0017] S12: Divide point clouds P and Q into segments perpendicular to the robot's forward direction with an appropriate width x. The intervals within each frame of the point cloud are denoted as A1, A2, A3, ... A j And B1, B2, B3, ... B k .

[0018] Further, step S2 includes:

[0019] S21: Calculate A1, A2, A3, ... A m And B1, B2, B3, ... B n Kurtosis of points within the interval;

[0020] The calculation formula is ;

[0021] S22: Calculate A1 and B1, B2, B3…B respectively. n A2 and B1, B2, B3…B n cosine value of the included angle ;

[0022] S23: Determine whether the cosine value of the angle between the point clouds of corresponding intervals in two frames is close to 1. If it is not close to 1, discard the point cloud of that interval; if it is close to 1, then the corresponding interval is the common interval of the point clouds of the two frames.

[0023] Further, in step S21, m4 is the fourth-order center distance of the samples, m2 is the second-order center distance of the samples, and x i For the i-th sample value, This is the sample mean.

[0024] Further, step S3 includes:

[0025] S31: Calculate the average Euclidean distance between points within the corresponding interval. The formula is as follows:

[0026] ;

[0027] S32: Euclidean distance The minimum case is the correct match of the nearest neighbor.

[0028] The present invention also provides a storage medium including a computer program, which, when executed, performs the above-described method for determining the nearest neighbor points of the inter-frame point cloud of the road scene to determine the nearest neighbor points of the inter-frame point cloud of the road scene.

[0029] The present invention also provides a robot, the robot comprising the above-described storage medium.

[0030] The method for determining nearest neighbor points between point clouds in road scenes of the present invention divides the point cloud into intervals, retains the common parts of the point cloud, eliminates the non-common parts, simplifies the calculation content, omits invalid calculations, and reduces the probability of erroneous corresponding points; it can improve the running efficiency and enhance the real-time performance of the point cloud matching algorithm without reducing the accuracy of subsequent matching. Attached Figure Description

[0031] Figure 1 This is a flowchart illustrating the method for determining the nearest neighbor point in a road scene point cloud according to an embodiment of the present invention. Detailed Implementation

[0032] The specific embodiments of the present invention will be described in further detail below with reference to the accompanying drawings and examples. The following examples are for illustrative purposes only and are not intended to limit the scope of the invention.

[0033] The terms "first," "second," "third," "fourth," etc., used in the specification and claims of this invention are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence.

[0034] Example 1

[0035] Figure 1 This is a flowchart illustrating the method for determining nearest neighbor points in inter-frame point clouds of a road scene according to an embodiment of the present invention. Please refer to... Figure 1 The method for determining nearest neighbor points in inter-frame point clouds of road scenes provided in this embodiment of the invention includes the following steps:

[0036] S1: Obtain point cloud data P and Q from the lidar two frames before and after time t, and perform directional segmentation on the point clouds in frames P and Q; specifically, step S1 includes:

[0037] S11: Name the two point cloud frames acquired by the lidar as set P and set Q;

[0038] S12: Divide point clouds P and Q into segments perpendicular to the robot's forward direction with an appropriate width x. The intervals within each frame of the point cloud are denoted as A1, A2, A3, ... A j And B1, B2, B3, ... B k .

[0039] S2: Remove non-common regions from the two point clouds;

[0040] Specifically, step S2 includes:

[0041] S21: Calculate A1, A2, A3, ... A m And B1, B2, B3, ... B n The kurtosis of points within the interval (the kurtosis of points is a feature of point cloud samples, and its purpose is to find corresponding point clusters of two sets of point clouds with similar features);

[0042] The calculation formula is ;

[0043] Where m4 is the fourth-order central distance of the sample, m2 is the second-order central distance of the sample, and x i For the i-th sample value, The mean of the sample;

[0044] S22: Calculate A1 and B1, B2, B3…B respectively. n A2 and B1, B2, B3…B n cosine value of the included angle ;

[0045] S23: Determine whether the cosine value of the angle between the point clouds of corresponding intervals in two frames is close to 1. If it is not close to 1, discard the point cloud of that interval; if it is close to 1, then the corresponding interval is the common interval of the point clouds of the two frames.

[0046] S3: Perform nearest neighbor matching on points within the corresponding interval of the common part;

[0047] Specifically, step S3 includes:

[0048] S31: Calculate the average Euclidean distance between points within the corresponding interval. The formula is as follows:

[0049] ;

[0050] S32: Euclidean distance The minimum case is the correct match of the nearest neighbor;

[0051] S4: Select the point corresponding to the case with the minimum average Euclidean distance, and the nearest neighboring point will be successfully determined.

[0052] It should be noted that due to the high speed of the robot moving in the road scene, a large amount of point cloud data is generated, and there are many non-common point clouds that are not needed in the nearest neighbor determination stage. Traditional methods for determining the nearest neighbor calculate the Euclidean distance for all points in two frames of point clouds, resulting in unnecessary computational costs. This invention proposes a method for determining the nearest neighbor between inter-frame point clouds in a road scene based on interval segmentation. It uses LiDAR to acquire inter-frame point cloud information of the road scene. By determining the one-to-one correspondence between corresponding points in two frames of point clouds, the preparatory work for subsequent point cloud registration is completed. In the process of determining the nearest neighbor, this invention uses an interval segmentation method to determine the inter-frame nearest neighbor. The point cloud data of two consecutive frames are input from the LiDAR, and the two frames of point clouds are directionally segmented to form intervals. By using the principle that the kurtosis of samples in the common interval is the same, the non-common parts are eliminated, the correspondence between intervals is determined, and the nearest neighbor is quickly determined within the corresponding interval.

[0053] Determining the nearest neighbor relationship between two point cloud frames quickly and accurately provides favorable conditions for subsequent registration. During robot movement, the onboard LiDAR generates point clouds by scanning the surrounding environment. Two consecutive frames of point cloud data are treated as two sets P and Q, where sets P and Q contain m and n LiDAR points, respectively. By segmenting the point cloud into intervals, the common parts of the point clouds are retained, while the non-common parts are eliminated, simplifying the calculations, omitting invalid calculations, and reducing the probability of incorrect point correspondences. This improves operational efficiency and enhances the real-time performance of the point cloud matching algorithm without reducing the accuracy of subsequent matching.

[0054] The present invention also provides a storage medium including a computer program, which, when executed, performs the above-described method for determining the nearest neighbor points of the inter-frame point cloud of the road scene to determine the nearest neighbor points of the inter-frame point cloud of the road scene.

[0055] The present invention also provides a robot, the robot comprising the above-described storage medium.

[0056] As can be seen from the above description, the advantages of this invention are:

[0057] 1. The method for determining the nearest neighbor point in the inter-frame point cloud of the road scene of the present invention requires the calculation of the average Euclidean distance of m×n cases when determining the nearest neighbor point, which is not suitable for situations with a large number of point clouds. In the present invention, the source point cloud and the target point cloud are divided into directional intervals, and the corresponding point cloud intervals are matched. The nearest neighbor point is calculated in the corresponding point cloud intervals, which simplifies the calculation content compared with the original method.

[0058] 2. The method for determining the nearest neighbor point in the road scene point cloud between frames of the present invention differs from the prior art, which calculates all points in the point clouds of two frames when determining the nearest neighbor point. Therefore, non-common parts are also included in the calculation steps during the process of determining the nearest neighbor point, resulting in invalid calculations. In the present invention, the non-common parts of the point clouds of two frames are distinguished by the directed interval segmentation method and are removed, thus avoiding the non-common part point cloud from participating in the calculation and omitting the invalid calculations in the original method.

[0059] 3. The method for determining nearest neighbors between point clouds in road scenes according to the present invention addresses the problem that existing technologies are prone to producing incorrect corresponding points when determining nearest neighbors in cases involving a large number of point clouds. In the present invention, by simplifying a large number of unnecessary calculations, the error rate for determining nearest neighbors is reduced, thereby improving the performance of the algorithm for determining nearest neighbors between point clouds.

[0060] 4. The method for determining nearest neighbors between point clouds in road scenes according to the present invention realizes the determination of nearest neighbors between point clouds in road scenes. It distinguishes the common and non-common parts of the point clouds between frames by using interval segmentation, reducing the amount of computation, and avoids invalid calculations by removing non-common parts of the inter-frames that are not needed in the nearest neighbor determination stage. Thus, it can quickly realize the determination of nearest neighbors between point clouds and improve the algorithm performance.

[0061] The above description is merely a specific embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the technical scope disclosed in the present invention should be included within the scope of protection of the present invention. Therefore, the scope of protection of the present invention should be determined by the scope of the claims.

Claims

1. A method for determining nearest neighbor points in inter-frame point clouds of a road scene, characterized in that, Includes the following steps: S1: Obtain point cloud data P and Q from the lidar before and after time t, and perform directional segmentation on the point clouds of frames P and Q. S11: Name the two point cloud frames acquired by the lidar as set P and set Q; S12: Divide point clouds P and Q into segments perpendicular to the robot's forward direction with an appropriate width x. The intervals within each frame of the point cloud are denoted as A1, A2, A3, ... A j And B1, B2, B3, ... B k ; S2: Remove non-common regions from the two point clouds; S21: Calculate A1, A2, A3, ... A m And B1, B2, B3, ... B n Kurtosis of points within the interval; The calculation formula is ; S22: Calculate A1 and B1, B2, B3…B respectively. n A2 and B1, B2, B3…B n cosine value of the included angle ; S23: Determine whether the cosine value of the angle between the point clouds of corresponding intervals in two frames is close to 1. If it is not close to 1, discard the point cloud of that interval; if it is close to 1, then the corresponding interval is the common interval of the point clouds of the two frames. S3: Perform nearest neighbor matching on points within the corresponding interval of the common part; S4: Select the point corresponding to the case with the minimum average Euclidean distance, and the nearest neighboring point will be successfully determined.

2. The method for determining nearest neighbor points in inter-frame point clouds of a road scene according to claim 1, characterized in that, In step S21, m4 is the fourth-order central distance of the samples, m2 is the second-order central distance of the samples, and x i For the i-th sample value, This is the sample mean.

3. The method for determining nearest neighbor points in inter-frame point clouds of a road scene according to claim 1, characterized in that, Step S3 includes: S31: Calculate the average Euclidean distance between points within the corresponding interval. The formula is as follows: ； S32: Euclidean distance The minimum case is the correct match of the nearest neighbor.

4. A storage medium, characterized in that, The system includes a computer program, which, when executed, performs the method for determining the nearest neighbor points of the inter-frame point cloud of a road scene as described in any one of claims 1 to 3.

5. A robot, characterized in that, The robot includes the storage medium described in claim 4.

Images