A vehicle tracking algorithm based on mean shift and C-V model

The traditional mean shift algorithm uses the tracking window of fixed size, so it is not efficient to track target vehicle, which continuously changes in size. To solve this problem, this paper presented a new vehicle tracking algorithm according to the characteristic of tracked vehicle. The new algorithm is based on the combination of mean shift and C-V model. Firstly, the initial tracking window was obtained with traditional mean shift tracking algorithm, then the tracking window was shifted and reshaped according to the shape information of target vehicle obtained by C-V model. In the tracking process, the shape and color information were both considered. Meanwhile, an improved C-V model was proposed, which used a new level set initialization function. Experiment results have shown that the tracking algorithm improves vehicle tracking accuracy greatly and guarantees real-time tracking.