A high precision target classification between different traffic participants that are of a complex composition, is the prerequisite for the follow-up tasks of autonomous driving such as planning decision. Millimeter-wave radar has drawn wide attention because of its robust capacity to observe all-weather and all day, but the traditional methods that are based on signal features has limitations, so they are not suitable for a direct application on autonomous driving. In this paper we have designed a traffic-participants-classification algorithm directed at the point cloud data that are attained with millimeter-wave radar. Besides, we carried out an experiment using high resolution 4-D millimeter-wave radar to collect our datasets, and then test the proposed algorithm, which contains manually extracted 14 features designed by us, with our data. Finally, the classification results are visualized and some comparisons are made. The proposed method can reach an average accuracy above 95% in 10-fold cross validation cross the 3 collected datasets, and the general performance better than the method in the reference.
An Algorithm of Millimeter-Wave Radar Target Classification Based on SVM Using Manually Extracted Point Cloud Features
Lect. Notes Electrical Eng.
International Conference on Intelligent Transportation Engineering ; 2021 ; Beijing, China October 29, 2021 - October 31, 2021
2021 6th International Conference on Intelligent Transportation Engineering (ICITE 2021) ; Chapter : 15 ; 170-177
2022-06-01
8 pages
Article/Chapter (Book)
Electronic Resource
English
Robust Multi-Target Tracking Algorithm Based on Automotive Millimeter-Wave Radar
SAE Technical Papers | 2018
|Robust Multi-Target Tracking Algorithm Based on Automotive Millimeter-Wave Radar
British Library Conference Proceedings | 2018
|Research on Tracking Algorithm for Forward Target-Vehicle Using Millimeter-Wave Radar
SAE Technical Papers | 2020
|Research on Tracking Algorithm for Forward Target-Vehicle Using Millimeter-Wave Radar
British Library Conference Proceedings | 2020
|