An algorithm for recognizing the markers in human motion detection is proposed in this paper. It includes three parts as follows: (1) an improved template matching method is used to acuminate the peak of the matching result. (2) The weighted average of the cross correlation between target image and template is given in order to make the resolution increase to sub-pixel level. (3) Forecasting the motion of marker and overcoming the hiding of marker are realized basing on the continuity of human motion. The result of experiments shows that the algorithm can recognize the positions of the markers at a relatively higher accuracy and speed. A gait analysis system based on this algorithm has come into use.