Autonomous navigation is one of the most essential capabilities of autonomous robots. In order to navigate autonomously, robots need to detect obstacles. While many approaches achieve good results tackling this problem with lidar sensor devices, vision based approaches are cheaper and richer solutions. This paper presents an algorithm for obstacle detection using a stereo camera pair that overcomes some of the limitations of the existing state of the art algorithms and performs better in many heterogeneous scenarios. We use both geometric and color based cues in order to improve its robustness. The contributions of the paper are improvements to the state of the art on single and multiple cue obstacle detection algorithms and a new heuristic method for merging its outputs.
Autonomous robots; Visual navegation; Obstacle detection