Hierarchical classification method for terrestrial laser point clouds

With the diversity of point clouds features between different objects and the irregularity of tree point clouds distribution, the lower classification accuracy of traditional method is a challenging problem. The hierarchical classification method for terrestrial laser scanning data is proposed. The non-ground points are segmented according to the Euclidean distance cluster. Scatter coefficient of normal presented by this paper is got. The trees are extracted by the scatter coefficient of normal. Combining with other features such as the mean of elevation, plane fitting residual, the number of points establishes a hierarchical classification method for point cloud. Experimental result show that the complex surface features can be classified. Compared with the density of projected point and SVM, the effectiveness of the proposed method has a demonstrable effect.