High resolution automotive radar data clustering with novel cluster method

Clustering of measurement data is an important task in digital signal processing. Especially in the case of radar signal processing the need of clustering detection points becomes obvious when high-resolution radar sensor systems are used. Clustering is usually used as a preprocessing step for classification of the measured data. In this paper a new approach for automotive radar data clustering is presented. A shape finding technique from image signal processing, called border following, is used to perform this task. Some adjustments and modifications of the method are required to get it working with radar measurements. The adapted algorithm is proven in three different measurement spaces and rated for the best performance by focusing on clustering of cyclists. It is showed, that the technique produces clustered radar data appropriate to their physical appearance.