Track Initiation Method Based on Improved Multi-density Adaptive DBSCAN

In the vehicle-mounted millimeter-wave radar scenarios, the traditional track initiation methods have problems such as low track initiation rate, a large number of false track initiations and track clustering. To solve above problems, an adaptive density-based spatial clustering of applications with noise (DBSCAN) parameter algorithm is introduced and improved in this paper, and a track initiation method based on improved multi-density adaptive DBSCAN is proposed. Firstly, the pre-processing technology, obtained by combining Hough transform and DBSCAN algorithm innovatively, can suppress the false track initiation from road edge clutters. Secondly, the pre-processed data is divided into different levels according to speed information, and DBSCAN parameters are adaptively generated by the median-based K-nearest neighbor method and mathematical expectation method, so as to improve the clustering effect of multi-density extended targets. Finally, an m/n logic is used to complete the track initiation. The simulation results show that the proposed method still has a track initiation rate of 85 % when the target scalability is 0.3. The analysis of measured data shows that the method can keep the number of false track initiation below 2 and the track initiation rate above 97 % in various road scenarios.