Application Effect of the Deep Neural Network PointNet in Ancient Architectural Carving in the Artificial Intelligence Environment

Abstract

Abstract: This study aims to better study the application of the deep neural network PointNet in ancient architectural carving and improve its efficiency of protection and restoration. Firstly, in the environment of artificial intelligence (AI) technology, a point cloud semantic segmentation model based on the deep learning neural network (DLNN) is established. Secondly, the PointNet point cloud semantic segmentation neural network is designed. Finally, the point cloud semantic segmentation neural network and the DLNN-based point cloud semantic segmentation model are verified experimentally in indoor and outdoor environments. The results show that the loss rate of the point cloud semantic segmentation model based on DLNN is finally stable at 0.05, the overall accuracy is 88.7%, and the mean intersection of union is 73.7%. The accuracy of the designed network model is above 95%. In the outdoor scene test, when the learning rate is 0.03, the overall model stability is the highest, the object roughness error is only 0.01, and the error of contour segmentation recognition is 0.398. These research data prove the feasibility and effectiveness of applying DNN PointNet in ancient architectural carving in the AI environment. This provides a new idea and method for protecting and restoring ancient architectural carvings.