A robot closed-loop detection method based on 3D point cloud

Abstract

The invention discloses a robot closed-loop detection method based on a three-dimensional point cloud. The method enables the robot to have the ability to identify scenes that have been reached. First of all, in view of the large amount of point cloud data and dense distribution in the 3D point cloud, point cloud voxel filtering is performed to down-sample the point cloud, and the number of point clouds is reduced and the algorithm is improved while keeping the original point cloud distribution characteristics unchanged. operation efficiency. At the same time, the PCA algorithm is used to process the down-sampled point cloud to determine the direction of the main axis, and to solve the problem of the impact of the robot rotation uncertainty on the closed-loop detection. Then, the point cloud processed by the PCA algorithm is subjected to dimensionality reduction processing, converted from the Cartesian coordinate system to the polar coordinate system, and rasterized into a two-dimensional image. Finally, the data structure Kdtree is used to speed up the search for candidate frames. In order to reduce the impact of discretization and noise on the two-dimensional image, a local column movement in the direction of the main axis is adopted to improve the efficiency of closed-loop detection and ensure the accuracy of closed-loop detection.

Description

technical field [0001] The invention belongs to the field of autonomous navigation of robots, and in particular relates to a closed-loop detection method of a robot's three-dimensional point cloud. Background technique [0002] Loop-closed detection, also known as loop-closed detection, refers to the ability of the robot to identify the scene it has arrived and make the map closed-loop, that is, the robot matches the scan of the current frame with the previously constructed map during the process of map building and positioning. If the loopback detection is successful, it can reduce the cumulative pose error of the robot composition, and the error detection will have a serious impact on the subsequent back-end optimization. Compared with visual closed-loop detection, 3D point cloud closed-loop detection has higher stability and is not affected by factors such as seasons, time, and lighting conditions. However, due to the lack of texture and color information and the large n...

AI Technical Summary

Problems solved by technology

However, due to the lack of texture and color information and the large number of point clouds, laser closed-loop detection also has certain difficulties.

At the same time, even if the robot returns to the same scene, due to the influence of the uncertainty of the heading rotation, the obtained 3D observation point cloud has different coordinate values ​​in the laser sensor coordinate system, which will also bring certain difficulties to closed-loop detection.

Images