A kind of adaptive multi-feature fusion tracking algorithm was proposed. The proposed algorithm overcame the particle degeneration phenomenon well by using finite-difference extended Kalman filter. The proposal distribution function was optimized. The latest observation information was fused into the suggestion distribution function by using finite-difference extended Kalman filter. Meanwhile, an adaptive multi-feature fusion method was proposed to overcome the defects of the additive fusion and the multiplicative fusion. The proposed method used static and dynamic characteristics as complementary observables in the framework of improved particle filter. Experimental results showed that the proposed method was effective in enhancing the accuracy and robustness of vehicle tracking system in different environments. 