Precise clustering of radar point clouds holds immense value in the context of training data annotations for various radar applications, including autonomous vehicles. However, due to the unique characteristics of radar data, such as sparsity, noise, and specularity, accurately separating radar detections into distinct objects poses a significant challenge. The traditional approaches of using location and Doppler as clustering features often fail when objects are in close spatial proximity and exhibit similar speeds: a scenario that is common in urban environments. To address this challenge, we introduce the concept of radar range–Doppler flow and a technique that extracts radial acceleration information of the surrounding targets. By incorporating radial acceleration into the feature space for radar point cloud clustering, we demonstrate a significant advantage over traditional methods, particularly when targets are in close proximity and move at similar speeds. Our approach provides an effective clustering solution in automotive radar applications in dense urban driving environments and any other similar situations where numerous targets coexist, and exhibit complex and unpredictable motion dynamics.
Radar Range–Doppler Flow: A Radar Signal Processing Technique to Enhance Radar Target Classification
IEEE Transactions on Aerospace and Electronic Systems ; 60 , 2 ; 1519-1529
2024-04-01
3140854 byte
Article (Journal)
Electronic Resource
English
GMM-Based Target Classification for Ground Surveillance Doppler Radar
Online Contents | 2006
|Signal processing for radar target analysis
Tema Archive | 1990
|