Objects classification from laser scanning data based on multi-class support vector machine

The classification of LiDAR point cloud is a key but difficult step for 3D reconstruction of architecture. The main classification methods are clustering-based and object-oriented. The support vector machine is an effective tactic which has been applied to classification, regression or other tasks. In this paper, we extract the vector angle, vector residual and position variance of point data as the key features of dimension value and put these key features into multi-class support vector machine, through calculating the probability of every point that belongs to each type, voting the maximum possible result. According to the voting result, we obtain the final classification result. The experiment results show that the classification method is promising.