A 3D Point Cloud Feature Identification Method Based on Improved Point Feature Histogram Descriptor

A significant amount of research has been conducted on the segmentation large-scale 3D point clouds. However, efficient cloud feature identification from results is an essential capability for computer vision and surveying tasks. Feature description methods are algorithms that convert set into vectors or matrices can be used identification. While histogram (PFH) descriptor method, it does not work well with objects have smooth surfaces, such as planar, spherical, cylindrical objects. This paper proposes a method based improved PFH feature-level normal efficiently distinguish surfaces. Firstly, established, then relationship between each point’s calculated. Finally, unknown identified by comparing similarity type-labeled feature. The proposed obtains overall accuracy ranging 71.9% to 81.9% street lamps, trees, buildings.