BASED ON IMPROVED POINTNET AUTOMATIC CLASSIFICATION METHOD OF GROTTO TEMPLE STATUES
Keywords: Grotto Temple, Point Cloud, Statue Classification, Improved PointNet, Automation
Abstract. As a carrier of history, culture, religion and art, grotto temples are an important part of China's splendid cultural heritage. Among them, there are many grotto statues with similar shapes and scattered on the grotto temples, which have high artistic value. It is of great significance to classify them.The existing classification methods of grotto temple statues are mainly based on traditional manual classification and machine learning classification, which often consumes a lot of labor costs and time costs. To solve these problems, this paper improves the PointNet network and applies it to the classification of statues in grottoes, which greatly improves the automation of the classification of statues. And the point cloud data set of the grotto temple is made for experiment. The results show that the overall accuracy of the method in this paper reaches 89.73%, the average intersection and combination ratio reaches 68.9%, and the accuracy is increased by 5.47% and 4.3% respectively compared with the random forest classification method. It is of great significance to the value research, status assessment and virtual restoration of the subsequent grotto temples.