摘要
光度立体三维(3D)重建是机器视觉和光度学领域中研究的热点问题,由于设备简单、成本低廉、分辨率高,得到广泛的应用。近年来,伴随着人工智能与深度学习技术的蓬勃发展,光度立体技术的发展进入一个崭新的时代。对深度学习技术在光度立体3D重建中的研究进展进行综述。首先,介绍光度学3D重建的研究背景和基本原理;其次,对光度立体3D重建方法的类型进行概述;接着,简要介绍常用的合成与实际拍摄数据集;然后,详细阐述深度学习技术在光度立体3D重建中的应用,它将基于物理模型的光度立体技术变为一种“数据驱动”下的技术,从而实现较高的预测精度;最后,进行分析与总结,并指出深度学习技术在光度立体领域所面临的挑战以及未来的研究趋势。
Photometric stereo three-dimensional(3D) reconstruction is a hot topic in the fields of machine vision and photometry.This method is widely used because of its simple equipment,low cost,and high resolution.With the rapid advancement of artificial intelligence and deep learning technology in recent years,the development of photometric stereo technology has entered a new era.This paper reviews the progress in the application of depth learning technology to photometric stereo 3D reconstruction.First,the research background and the basic principles of photometric 3D reconstruction are introduced.Next,various types of photometric stereo 3D reconstruction methods are summarized.Then,the commonly used synthetic and real-photoed datasets are briefly introduced.Further,a detailed description of the applications of depth learning technology in photometric stereo 3D reconstruction is provided,wherein the physical modelbased photometric stereo technology is transformed into a “data-driven” technology to achieve high prediction accuracy.Finally,the paper analyzes and summarizes the challenges and opportunities for future research in the application of deep learning technology to photometric stereo reconstruction.
作者
王国珲
卢彦汀
Wang Guohui;Lu Yanting(School of Optoelectronic Engineering,Xi’an Technological University,Xi’an 710021,Shaanxi,China)
出处
《激光与光电子学进展》
CSCD
北大核心
2023年第8期187-206,共20页
Laser & Optoelectronics Progress
基金
陕西省教育厅科学研究计划项目资助(22JY025)
陕西省自然科学基础研究计划资助项目(2022JM-318)。
关键词
光度立体
三维重建
光度学
人工智能
深度学习
数据驱动
photometric stereo
three-dimensional reconstruction
photometry
artificial intelligence
deep learning
data-driven
