Advantage of Different Data Dimensions and Dual-Beam for Radar Low Resolution Doppler Classification

Wind farms and other challenging environments in combination with more sensitive radar systems lead to more and more clutter detections. This significantly affects the operational usability of the radar display. In the past, we have shown that micro-Doppler based target classification can be used to distinguish different targets as well as clutter from each other even for low resolution Doppler radars (LRDRs). Here, we investigate two approaches to further increase the classification performance by adding different data dimensions. We will first present the in-house dataset from an air surveillance radar, associated challenges, and the model used to perform the target classification. We will then discuss the results and present an ablation study comparing the performance between range-Doppler information only, and the full range-Doppler-azimuth cuboid as input, using single and dual-beam data. It is shown, that the additional information further increases classification performance, improving operational experience.