Comparative Study of Unsupervised Clustering Methods Used for RADAR Applications

Driver safety has become an important aspect. To have driver safety RADAR is an essential part of vehicles hence RADAR has great significance in the automotive industry. The Radar sensor collects data from surroundings that may have unwanted data that may lead to improper detections of intended objects, so to have proper object detections it is needed to use clustering methods on the radar point cloud data. There are numerous unsupervised clustering methods used for RADAR applications. In this paper, the comparisons of different unsupervised algorithms such as K-Means Clustering, Hierarchical Clustering, Cluster Using the Gaussian Mixture Model, and DBSCAN are presented. All these clustering algorithms are evaluated based on various evaluation criteria such as the Silhouette coefficient, Davies Bouldin index, etc. Based on evaluations and comparative studies applications of the clustering algorithms are classified.