A Novel Possibilistic Clustering Algorithm for Measurement Data of Vehicle MMW Radar

The performance of measuring moving vehicle targets with onboard millimeter-wave (MMW) radar can be affected by environmental noise, as well as the dynamic changes in the number of these moving vehicles and the problem of interference of measurement data between adjacent vehicles. In this article, we propose a new possibilistic clustering algorithm for this problem. The radar measurement dataset will contain a lot of information about stationary targets prior to cluster analysis. To obtain a moving target measurement dataset, a method for recognizing stationary and moving target measurements was developed. Then, based on the moving vehicle target measurement dataset, it is initialized with an improved fast density peaks (IFDPs) clustering algorithm, and an enhanced adaptive possibilistic c-means (EAPCM) clustering algorithm is implemented. Because the EAPCM algorithm has the ability to delete redundant clusters but does not increase the clusters, the number of clusters obtained by the IFDP algorithm will be greater than that obtained by the fast density peak (FDP) algorithm. Furthermore, the EAPCM algorithm improved the clustering center point of the adaptive probabilistic c-Means (APCM) algorithm, and replaced it with an adaptive line segment, which makes the algorithm adapt to the vehicle-mounted MMW radar measurement dataset. Two different experiments are carried out in this article, the results show the stability and reliability of the stationary and moving vehicle target recognition method in extracting the measurement values of moving targets, and when compared to other algorithms, the EAPCM algorithm has a higher classification rate (CR) and matching accuracy.