摘要
三维重建在自动驾驶、元宇宙等工业领域有着重要的应用前景。其中,透明刚体的三维重建由于其复杂的成像规律在重建过程中存在一定挑战。本综述回顾了透明刚体的三维重建技术,重点介绍了基于优化可视外壳的方法和基于深度学习的方法,讨论了基于可视外壳重建的原理和各类优化方法,还讨论了基于深度学习的RGB-D深度补全和多视图方法,并介绍了目前透明物体的数据集,最后对可能的发展方向进行了讨论和展望。
The three-dimensional(3D)reconstruction technology has considerable application prospects in industrial fields such as autopilot and metauniverse.The 3D reconstruction of transparent rigid bodies exhibits some challenges owing to its complex imaging laws.This review examines the 3D reconstruction technology of transparent rigid bodies,focuses on the methods based on optimized visual hull and deep learning,discusses the principle and different optimization methods based on visual hull reconstruction,describes the RGB-D depth completion and multiview reconstruction based on deep learning,presents the current dataset of transparent objects,and finally discusses the possible development direction of the 3D reconstruction technology.
作者
潘智辉
陈睿
胡昌平
徐静
Pan Chifai;Chen Rui;Hu Changping;Xu Jing(Department of Mechanical Engineering,Tsinghua University,Beijing 100083,China)
出处
《激光与光电子学进展》
CSCD
北大核心
2023年第8期86-94,共9页
Laser & Optoelectronics Progress
关键词
三维重建
透明物体
可视外壳
深度学习
three-dimensional reconstruction
transparent object
visual hull
deep learning
