One method is presented in the paper to perform multi-pedestrian tracking in infrared video surveillance,based on the moving region in the images.Firstly,the background model is built by the 3D median filter,and the background-subtraction method is used to detect moving regions.Secondly,one matrix is created between two sequential frames using the match criterion between moving regions and finally the tracking is realized depending on the matrix.According to the test result by infrared image series,the method is robust and can track pedestrians fast and effectively.