Distance Estimation of Preceding Vehicle Based on Mono Vision Camera and Artificial Neural Networks

Most of the research in automotive literature has been focused on the object detection task because of which distance estimation between vehicles has gained less attention in the computer vision. Distance between vehicles is foremost important to prevent accidents and enhance vehicle safety. To keep a safe distance between vehicles and avoid the traffic accidents, traditional method of inverse perspective mapping algorithm is used. It is observed that the algorithm resulted in poor performance for far distance objects. To solve the problem of high error rate of preceding vehicle distance measurement of driver assistant system, this paper proposes an Artificial Neural Network (ANN) based algorithm, which is used to calculate the distance of preceding vehicle from camera images. The proposed model uses a supervised learning algorithm, YOLO (you look only once) object detector for vehicle detection. The input and output attributes were evaluated from the object bounding boxes, which is obtained from the object detector. The proposed ANN based system was estimated on the straight road images. The experiments conducted on KITTI benchmark, indicate that the proposed method has reduced the error rate to 2% compared with other existing methods.