3D Point Cloud Generation with Millimeter-Wave Radar
summary
System converts commodity vehicle Radar into coherent SAR that outputs 3D point clouds by combining self cross-range tracking with a multi-focusing step and elevation estimation across vertical antenna pairs. Implementation runs on a TI cascade radar with 6 transmit, 8 receive antennas and uses a ZED stereo camera for ground truth. MilliPoint achieves wavelength-scale localization of the radar, resolves specular reflections by detecting focusing centers in the spatial frequency spectrum, and produces much denser, higher-resolution point clouds of cars, pedestrians, bikes and bins compared with noncoherent and static-radar baselines.
AI-written summary · about
related talks
21mRadarCat: Radar Categorization for Input & Interaction
16m"Putting a New Spin on
24m3D Backscatter Localization for Fine-Grained Robotics
26mFrom Panoramic Vision to Lightweight 3D Worlds for Semantic Visual Tracking
28mThe Role of High Definition 3D LiDAR in Self Driving Cars
20mCarMap: Fast 3D Feature Map Updates for Automobiles
21mThe race to replace human drivers
41mComputer Vision for Remote AR
