The ability to detect obstacles is essential for safe vehicle guidance. In this paper, we present a real-time low-cost obstacle detection system using a pair of cameras. With the assumption of a ground plane, IPM (Inverse Perspective Mapping) is first used to remove the perspective effect in the original images. Free-space in front of the vehicle is then detected by inspecting the difference between the two ground images. After a filter is applied to remove the noise in the free-space image, obstacles are located using two histograms. This system has been tested on THMR-V (Tsinghua Mobile Robot V), an intelligent vehicle developed at Tsinghua University. Through various experiments we successfully demonstrate its real-time performance and robustness with respect to different road textures.
展开▼