Optical Flow Based Urban Road Vehicle Tracking
Liu Ya, Yao Lu, Qingxuan Shi, Jianhua Ding · 2013
Vehicle tracking is an important part in intelligent transportation surveillance. But now vehicle tracking faces with the problems such as scale change, the interference of similar color, low resolution video data and so on. In this paper an improved Markov chain Monte Carlo(MCMC) named optical flow MCMC(OF-MCMC) sampling tracking algorithm is proposed for vehicle tracking. First, we use the optical flow method to get the moving direction of the vehicle in initial frames, which can solve the problem of scale change, what's more the optical flow method can get the moving speed of the vehicle which replaces the second-order autoregressive motion model owing to the non-parameter characteristic. Second, when calculating whether one particle is accepted or not, a distance factor is considered, which can relieve the interference of similar vehicle nearby. Finally, to deal with vehicle tracking in low resolution of the video data, we generate a more accurate feature template with different features weighted to get better tracking results. Experimental results show that the proposed tracking algorithm has better performance than some traditional ones.