Road sides recognition in non-structured environments by vision

The present communication deals with a vision system designed to detect and track roadsides on non-marked roads or paths. The initial detection method, mainly adapted to marked roads and briefly presented in this article, is based upon a model driven algorithm. This article presents the major improvements we achieved to adapt the approach to the particular context of non marked roads. These improvements are based on the development of a pre-processing step to improve the quality of the images, the transition detector, the recognition criterion. Moreover, the new approach is able to take into account the proprioceptive data provided by an inertial measurement unit. Results obtained show the robustness and the accuracy of the method.