Virtual Multiview Fusion for Millimeter Wave Radar Point Cloud Generation

Abstract

Conventional millimeter wave (mmwave) point cloud generation technology suffers from information loss due to sparse scattering points on targets. Existing works generate and fuse radar data to enhance the point cloud, but they either demand datasets or consume extra resources. This letter proposes a virtual multiview fusion system for mmwave point cloud generation to attain complete target characteristics with the least resources. In our system, we set a single radar for sensing and regard radar signals relying on walls as virtual detection from multiple views. Then, we fuse target features detected from virtual views to the direct path detection to densify the point cloud. Instead of mitigation, multipath components are reserved and employed as supplements. It contains new characteristics from different perspectives, effectively compensating for the specular reflection loss without additional detection. Experiments are performed to validate the effectiveness of the proposed system in generating a dense radar point cloud.