Point Cloud Classification Algorithm Based on the Fusion of the Local Binary Pattern Features and Structural Features of Voxels

Point cloud classification is a key technology for point applications and feature extraction step towards achieving classification. Although there are many methods, the acquisition of colored data has become easier in recent years, most processing algorithms do not consider color information associated with or make full use information. Therefore, we propose voxel-based local descriptor according to binary pattern (VLBP) fuses RGB geometric structure features using random forest classifier build algorithm. The proposed algorithm voxelizes cloud; divides neighborhood center into cubes (i.e., multiple adjacent sub-voxels); compares gray voxel sub-voxels; performs global thresholding convert it code; uses difference sign–magnitude transform (LDSMT) decompose an entire two complementary components sign magnitude. Then, VLBP each extracted. To obtain more structural about cloud, method extracts normal vector corresponding fast histogram (FPFH) based on vector. Finally, mechanism (normal FPFH) (RGB features) fused, used classify laser cloud. experimental results show that can achieve effective from different indoor outdoor scenes, improve accuracy