It is difficult to track the moving vehicle due to various factors including complex environment, changes of illumination and scale. A moving vehicle tracking algorithm based on deep learning is proposed in this paper. First of all, traditional GMM algorithm is improved to reduce the error judgment probability of pixel state. Then, a sparse DAE neural network feature learning framework is proposed to ensure efficient extraction of vehicle features and reduce feature redundancy. Finally, the vehicle is tracked in its area. The experimental results show that the CM of proposed algorithm achieves 0.85 and has strong robustness.