Automatic extraction method of independent features based on elevation projection of point clouds and morphological characters of ground object

In order to extract independent features automatically from vehicle-borne laser point clouds, this paper, focused on poles and trees, develops an automatic extraction method based on elevation projection of point clouds and morphological characters of ground object. Firstly, random point clouds are transformed into a digital image by elevation projection with the gray degree of the image representing height. Secondly, a series of digital image processing techniques, such as binarization, morphological dilation, trim, are used to highlight the pixels of independent features and then abstract them. Thirdly, pixels with 2D coordinates are reversely calculated into several sets of cloud points with 3D coordinates according to the coordinate transformation relationship between digital image and point clouds, thus segmentation of independent features can be finished. Finally, the classification based on point cloud sets of independent features can be achieved according to the morphological characteristics of independent features. The experiment result shows that the proposed method in automatic extraction of independent features is feasible, and the accuracy of the classification can be as much as 92%.