A Panoramic Color Vision System for Following Ill-Structured Roads

The ability to follow ill-structured roads, such as footpaths, dirt tracks and corridors, is important for mobile robot navigation. This paper presents a panoramic colour vision based road following system for ill-structured roads. Roads are modeled with rapidly adapting 3D colour histograms and a simple yet generic geometric model. The computational complexity of the geometric model fitting stage has been significantly reduced compared to other works. A Kalman filter is used to smooth out any measurement noise. Results from experiments in tracking a footpath demonstrate the robustness of the system.