A smart moving vehicle detection system using motion vectors and generic line features

Abstract

This paper presents a novel algorithm for the detection and tracking of low relative speed moving vehicles. The proposed algorithm is particularly suitable for mass-produced in-vehicle devices as it combines with motion vector based moving object detection to form a complete solution for an Advanced Driver Assistance System given its reduction in cost and complexity. The algorithm utilizes motion vectors that are readily available from video encoder output. The region of interest for detection is reduced by ignoring the area above the vanishing line of the captured image, evaluation of the amplitude of motion vectors and identification of the road region. During the evaluation process, a binary image is generated by comparing the gray-level of the captured image to the gray-level of the detected road region. Subsequently, the horizontal and vertical contours of specific areas inside the region of interest are evaluated. Test results show the effectiveness of the algorithm with more than 90% detection rate and the suitability for real-time use with cycle time of less than 66ms.