Classification Algorithm for Laser Point Clouds of High-steep Slopes Based on Multi-scale Dimensionality Features and SVM

In order to solve the vegetation filtering problem of high-steep slope point cloud data in complex scene, the multi-scale dimensionality feature of vegetation and rock laser point cloud on high-steep slope is studied first. Then the SVM (support vector machine is utilized) to build a classifier. Finally a vegetation filtering algorithm of high-steep slope laser point cloud is proposed and a three-dimensional laser point cloud filtering software LIDARVIEW is written . The data shows that: the vegetation of different scales in complex scene is well identified and the classification accuracy of the filtering algorithm is high. The algorithm is not affected by the density, occlusion of laser point cloud as well as the complex topography and it is also suitable for airborne LiDAR point cloud data filtering. The classification accuracy of rock under high vegetation cover is greater than 93%, while under low vegetation cover is higher than 97%. The algorithm has great significance for hilly high-steep slope terrain measurement with complex topography.