A 4D millimeter-wave radar point cloud processing and target classification method

Abstract

The invention discloses a 4D millimeter-wave radar point cloud processing and target classification method. The method includes point trace input, point trace preprocessing, Kalman filter prediction, point trace and track association, point trace clustering, track start, track update and track management. The invention realizes the transition from a two-dimensional plane to a three-dimensional plane, and the dot trace characteristics of the target are more obvious. Construct the target measurement virtual point track in a SNR weighted way, project the associated point track onto the xoy plane, rotate the course angle of the xoy plane clockwise with the virtual point track as the origin, and use the multi-frame sliding window method to The positional relationship displacement between virtual points calculates the size information of the target, which improves the unobvious feature of the target size due to the sparseness of the millimeter-wave point cloud. According to the characteristics of the target, the class probability of the single-frame target is given, combined with the weighted form of historical probability and single-frame probability. The target classification probability of this frame, the class with the highest probability in this frame is used as the final classification result of this frame.

Description

technical field [0001] The present invention relates to the technical field of 4D millimeter-wave radar point cloud processing and target classification methods, in particular to a 4D millimeter-wave radar point cloud processing and target classification method. Background technique [0002] At present, due to the sparseness of the point cloud of the millimeter-wave radar, the features of the target contained are few, and the impact of different orientations of the target on the target classification during the driving process is not fully considered in the classification method using the point trace feature, resulting in the The target classification method has low accuracy and poor practicability, so it brings great challenges to the development of millimeter-wave radar in practical applications. The existing target classification based on millimeter-wave radar is mainly used to distinguish pedestrians and vehicles, while the market The need for object classification doesn...

AI Technical Summary

Problems solved by technology

[0002] At present, due to the sparseness of the point cloud of the millimeter-wave radar, the features of the target contained are few, and the impact of different orientations of the target on the target classification during the driving process is not fully considered in the classification method using the point trace feature, resulting in the The target classification method has low accuracy and poor practicability, so it brings great challenges to the development of millimeter-wave radar in practical applications. The existing target classification based on millimeter-wave radar is mainly used to distinguish pedestrians and vehicles, while the market The need for object classification goes beyond that

Images