Point cloud classification method based on point cloud semantic annotation and optimization

• Main Authors: HUANG RONG, GU ZHENXIONG, XU YUSHENG, PAN YUE, YE ZHEN
• Format: Patent
• Language: Chinese; English
• Published: 06.09.2019
• Subjects: CALCULATING; COMPUTING; COUNTING; HANDLING RECORD CARRIERS; IMAGE DATA PROCESSING OR GENERATION, IN GENERAL; PHYSICS; PRESENTATION OF DATA; RECOGNITION OF DATA; RECORD CARRIERS

The invention relates to a point cloud classification method based on point cloud semantic annotation and optimization, and the method comprises the following steps: 1, carrying out the pre-classification of original point cloud data through employing PointNet++, and obtaining a point cloud pre-classification result; and 2, carrying out classification result optimization on the pre-classificationresult by using global space regularization to obtain a final point cloud classification result. Compared with the prior art, the invention provides the universal framework for obtaining the point cloud semantic tags and improving the classification result. In a proposed universal framework, the existing steps can be replaced by using a similar algorithm, and an optimization method based on graphstructure regularization is carried out on an initial annotation result of the three-dimensional point cloud, so that the spatial smoothness of semantic annotation is realized, and only a small amountof training data is requir