Abstract: Mathematical morphology is widely used in the digital image processing. In this paper, the characteristics of traditional mathematical morphology algorithm are introduced and the shortages of the application of this theory in the LIDAR point clouds data classification and filtering are analyzed first. Then, on this basis, the corresponding morphological algorithms are efficiently improved and extended, and an adaptive filtering algorithm which is aimed at the characteristics of different terrain surfaces is proposed. Based on the “opening” operator of mathematical morphological, a method that a “bandwidth” parameter is added for the classification and filtering of LIDAR point clouds is proposed. Finally, experiments are processed using three groups of actual LIDAR point clouds data and the validity of the algorithm is validated by the different presentation forms from them.