An adaptive directional filter for photon counting Lidar point cloud data

An adaptive directional filter method was proposed for the photon counting Lidar point cloud data. The method defines a filter kernel with its main filter direction adjustable. The density of the best filter direction were achieved by traverse and the noise points away from the objects were removed. The noise points adjacent to the objects were eliminated according to the density difference between the point and points in its neighborhood. The filtering method provide here is validated through the point cloud data obtained in an aerial experiment. The results show that the filtering method is able to eliminate the noise points very close to the ground effectively and is fit for the low density object point cloud recognition, while the filter accuracy is 91.86% for vegetable points and 97.89% for ground points.