Mean Shift algorithm works well locally and is used to track the objects with low speed.But the algorithm is easy to fall into local maximum due to lack of appropriate target model updating mechanism.In this paper, a new Mean Shift algorithm based on SIFT Verification is proposed, which can be used for tracking the occluded object moving at high speed.The proposed algorithm can compensate for the shortcoming of the Mean Shift while keeping fine real-time performance.Tests were made to the scenes of occluded vehicle moving at high speed with comparison to the traditional algorithms.The experimental results show that the proposed algorithm can effectively track an object under the complex condition with fine real-time performance and high robustness.