Abstract: An automatic, real-time no-rigid detection and tracking algorithm is presented. We detect moving objects by use the dynamic background model algorithm. Color histogram is not affected by object's scale and has good robustness, and mean shift algorithm is a nonparametric density gradient estimator. The robustness and speed is improved by integrating a mean-shift based model update technique with an adaptive change detection method. The result of the experiment shows the method is practical and efficient.