PROCEDE DE DETECTION D'OBSTACLES DEPUIS UN VEHICULE AUTOMOBILE

摘要

The method involves selecting points of interest on an acquired image, and calculating a speed vector associated to each selected point of interest. The speed vectors oriented towards the center of a following image are conserved. Zones of the acquired image corresponding to the origin of the conserved vectors are magnified to constitute interest zones. The constituted zones are temporally averaged on consecutive following images. Closed contour blobs are temporally identified among the averaged interest zones, and identified blobs are compared to obstacles for a considered motor vehicle.