Abstract

Obstacle detection is a main key of autonomous vehicles. When communicating with huge robots in unstructured background, resilient obstacle detection is required. Few of the existing methods are mainly suited for the backgrounds in which the ground is comparatively flat and with roughly the same color throughout the terrain. A novel procedure proposed in the work presented here uses a monocular camera, for real-time performance. We compute the homographic between two successive frames by computing the fundamental matrix between the two frames. Estimation of fundamental matrix is followed by triangulation so as to estimate the distance of the object from the camera. We examine a difficulty intrinsic to any fundamental matrix-based outlook to the provided task, and show how the discussed way can resolve this difficulty by a huge level. An obstacle detection and distance estimation system based on visual particular attribute and stereo vision is hence discussed in the presented work.

Authors and Affiliations