Monocular vision-based collision avoidance system using shadow detection

This research paper is devoted for discussing a Vision-Based Collision Avoidance System that aims to provide the driver with a “third eye” to help him/her to detect obstacles and estimate distance between them and the host vehicles. It is based on a monocular approach of image processing that has one camera, which continuously captures images of the frontal view of the vehicle. Then the captured images are processed in order to detect obstacles, then estimate their distances from the host vehicle and, finally, take decisions to avoid them. The detection algorithm depends on detecting the shadow of the obstacles, as an invariant feature for all types of obstacles. Watershed segmentation technique is used to detect objects and triangulation technique is used to calculate the distance between the host vehicle and the detected obstacle. The proposed system can automatically control electric vehicles.