Imaging millimeter wave radar point cloud target classification method based on machine learning

Abstract

The invention relates to an imaging millimeter-wave radar point cloud target classification method based on machine learning, and the method comprises the following steps: S1, obtaining an input point cloud of a millimeter-wave radar, preprocessing the input point cloud, obtaining a detected target point cloud cluster, and constructing a target point cloud data set according to the detected target point cloud cluster; S2, generating a minimum cubic envelope frame of the target point cloud, calculating and extracting feature parameters of the target point cloud, and generating a feature vector of target classification; and S3, inputting the feature vector into a pre-trained classification model to obtain a target type corresponding to the detected target point cluster. Compared with the prior art, the method has the advantages that the practicability is high, various types of target classification can be realized, and the target classification capability of the millimeter wave radar is greatly enhanced.

Description

technical field [0001] The invention relates to the field of intelligent driving environment perception and millimeter wave radar target classification, in particular to a machine learning-based imaging millimeter wave radar point cloud target classification method. Background technique [0002] With the rapid development of artificial intelligence and intelligent equipment, smart cities and intelligent driving technology have also made great progress. Millimeter-wave radar has been widely used in the fields of intelligent driving, intelligent transportation and intelligent security because of its advantages of all-weather and full-time operation and powerful ranging and speed measuring capabilities. Although we can obtain the position and speed information of the target through radar, the mature millimeter-wave radar products in the current market cannot accurately identify and distinguish the type of target, which limits the application scenarios of radar. [0003] The ex...

AI Technical Summary

Problems solved by technology

The above methods have low classification accuracy, few types that can be classified, and insufficient real-time and robustness of the algorithm, so they cannot be directly promoted in practical applications.

[0004] In addition, with the development of millimeter-wave radar products, the new generation of imaging millimeter-wave radar adopts technologies such as multi-transmission and multi-reception, multi-antenna array cascading, beamforming, and ultra-wideband to further improve the range and angle resolution of the radar. The height information of the target, the existing technology cannot be applied to this kind of imaging millimeter wave radar for target classification

[0005] According to the Chinese patent CN201910640745.7, a nuclear support vector machine target classification method based on millimeter-wave radar point cloud features is disclosed. This patent designs an 11-dimensional feature vector for the traditional two-dimensional radar point cloud. The feature vector does not consider radar points. The height information of the cloud and the distribution of point cloud reflection intensity cannot be applied to new imaging millimeter-wave radar products. This patent only completes the classification of people and vehicles, and cannot cover most types of targets in road traffic.

Images