Accurate vehicle location and tracking algorithms based on improved kernelized correlation motion model and Kalman filter in intelligent transport surveillance system

Abstract

Vision-based traffic surveillance, an indispensable part of intelligent transport system (ITS), has been widely studied over past few years. Due to the factors such as visual occlusion, illumination change and pose variation, it is a challenging task to develop effective and efficient models for vehicle detection and tracking in surveillance videos. Although plenty of existing related models have been proposed, many problems still need to be resolved. A novel improved kernelized correlation tracking algorithm based on Kalman filter and motion model is proposed to solve the tracking accuracy deteriorated by fast-moving and severe occlusion. Firstly, an adaptive search region location method is proposed, where the optimal position is estimated by the uncertainty theory of the motion model so as to obtain the optimal search window. Secondly, a confidence strategy for similarity measurement is calculated by using the sharpness of the correlation peak and the smoothness constraint. Finally, the optimal confidence is introduced to get an adaptive update model. A large number of simulation experiments show that our proposed algorithm has more robust performance and anti-interference ability than the traditional KCF algorithm, which is suitable for intelligent transport system application.