基于Point Net++的碎片断裂面提取算法  

Fragment Fracture Surface Extraction Algorithm Based on PointNet

作　　者：刘旭斌 刘嵩[1] 刘建成[2] 王冲 LIU Xubin;LIU Song;LIU Jiancheng;WANG Chong(Key Laboratory of Information and Computing Science of Guizhou Province,Guizhou Normal University,Guiyang,Guizhou 550001,China;Sichuan Provincial Cultural Relics and Archeology Research Institute,Chengdu,Sichuan 610041,China)

机构地区：[1]贵州师范大学贵州省信息与计算科学重点实验室,贵州贵阳550001 [2]四川省文物考古研究院,四川成都610041

出　　处：《自动化应用》2024年第6期129-132,135,共5页Automation Application

摘　　要：针对文化遗产修复问题,提出了一种基于深度学习的碎片断裂面提取的算法,通过基于PointNet++的神经网络结构对碎片的断裂面进行了提取,并在FragTag3D数据集与3DPuzzle数据集上进行了测试。结果表明,该方法能够提取这些数据集中大部分碎片的断裂面。A deep learning based algorithm for extracting fragmented fracture surfaces was proposed to address the issue of cultural heritage restoration.The fracture surfaces of fragments were extracted using a neural network structure based on PointNet++,and tested on the FragTag3D dataset and 3DPuzzle dataset.The results indicate that this method can extract the fracture surfaces of most fragments in these datasets.

关 键 词：深度学习 PointNet++ 语义分割 碎片拼接 断裂面提取 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]