3D计算机视觉：原理、算法及应用

图像处理领域专家章毓晋教授全新力作；
理论讲解 应用示例；
内容全面，讲解透彻，全篇干货。

本书主要内容围绕3D 计算机视觉展开，介绍了相关的基础概念、基本原理、典型算法、实用技术和应用成果。本书可在学过其姊妹篇《2D 计算机视觉：原理、算法及应用》后学习。 本书将从客观场景出发到后对场景进行理解的全过程分为5 个部分进行介绍。第1 部分是图像采集，介绍了摄像机标定和3D 图像采集技术；第2 部分是视频运动，介绍了视频图像和运动信息，以及对运动目标进行检测和跟踪的技术；第3 部分是物体重建，介绍了双目立体视觉和单目图像恢复技术；第4 部分是物体分析，介绍了3D 目标表达和广义匹配；第5 部分是高层理解，介绍了知识和场景解释及时空行为理解。本书除提供大量示例外，还针对每章的内容提供了自我检测题（含提示并附有答案），并且给出了相关的参考文献和术语索引（包括英文）。

章毓晋，于1989年获比利时列日大学应用科学博士学位。1989—1993年，先后在荷兰德尔夫特大学从事博士后研究工作并担任专职研究员。于1993年到清华大学任教，1997年被聘为教授，1998年被聘为博士生导师，2014年成为教学科研系列长聘教授。在2003年学术休假期间，同时被聘为新加坡南洋理工大学访问教授。在清华大学，先后开出并讲授10多门本科生和研究生课程。在南洋理工大学，开出并讲授研究生课程“现代图像分析（英语）”。已编写出版了图像工程系列教材第1版、第2版、第3版和第4版，《图像工程问题解析》、《图像处理和分析基础》、《图像处理和分析技术》（第2版和第3版）、《图像处理和分析教程》（第1版、第2版和第3版）、《计算机视觉教程》（第1版、第2版和第3版）和《图像处理基础教程》，以及Image Engineering: Processing, Analysis, and Understanding，Image Engineering, Vol.1, Image Processing，Image Engineering, Vol.2, Image Analysis，Image Engineering, Vol.3, Image Understanding，翻译出版了《彩色数字图像处理》、《图像处理基础》（第2版）、《MATLAB图像和视频处理》、《计算机视觉基础》和《彩色计算机视觉》，研制出版了《“图像处理和分析’’多媒体计算机辅助教学课件》和《图像处理和分析网络课程》。已在国内外发表了30多篇教学研究论文。主要科学研究领域为所提出的图像工程（图像处理、图像分析、图像理解及其技术应用）。自1996年起，已连续26年对中国图像工程的研究及主要文献进行了系统的年度分类和总结综述。已在国内外发表了500多篇图像工程研究论文，出版了专著《图象分割》《基于内容的视觉信息检索》《基于子空间的人脸识别》，编著了《英汉图像工程辞典》（第1版、第2版和第3版）、《图像工程技术选编》和《图像工程技术选编（二）》，主持编著了Advances in Image and Video Segmentation，Semantic-Based Visual Information Retrieval，Advances in Face Image Analysis: Techniques and Technologies，出版了Handbook of Image Engineering和A Selection of Image Processing Techniques.曾任第24届国际图像处理学术会议（ICIP‘2017）等20多个国内外学术会议的程序委员会主席。现为中国图象图形学学会名誉监事长和会士、国际电气电子工程师协会（IEEE）高级会员、国际光学工程协会（SPIE）会士（因在图像工程方面的成就）。

第1 章计算机视觉概述·····································································.1
1.1 人类视觉及特性····································································.1
1.1.1 视觉特点····································································.2
1.1.2 视觉的亮度特性···························································.3
1.1.3 视觉的空间特性···························································.5
1.1.4 视觉的时间特性···························································.6
1.1.5 视知觉·······································································.8
1.2 计算机视觉理论和框架··························································.10
1.2.1 计算机视觉的研究目的、任务和方法·······························.11
1.2.2 视觉计算理论·····························································.11
1.2.3 框架问题和改进··························································.16
1.3 3D 视觉系统和图像技术·························································.18
1.3.1 3D 视觉系统流程·························································.18
1.3.2 计算机视觉和图像技术层次···········································.19
1.3.3 图像技术类别·····························································.20
1.4 本书结构框架和内容概况·······················································.21
1.4.1 结构框架和主要内容····················································.22
1.4.2 各章概况···································································.22
1.5 各节要点和进一步参考··························································.23
第2 章摄像机标定···········································································25
2.1 线性摄像机模型···································································.25
2.1.1 完整成像模型·····························································.26
2.1.2 基本标定程序·····························································.27
2.1.3 内、外参数································································.28
2.2 非线性摄像机模型································································.30
2.2.1 畸变类型···································································.31
2.2.2 标定步骤···································································.33
2.2.3 标定方法分类·····························································.34
2.3 传统标定方法······································································.35
2.3.1 基本步骤和参数··························································.36
2.3.2 两级标定法································································.36
2.3.3 精度提升···································································.40
2.4 自标定方法·········································································.41
2.5 各节要点和进一步参考··························································.44
第3 章 3D 图像采集··········································································45
3.1 高维图像············································································.45
3.2 深度图···············································································.46
3.2.1 深度图和灰度图像·······················································.47
3.2.2 本征图像和非本征图像·················································.47
3.2.3 深度成像方式·····························································.49
3.3 直接深度成像······································································.50
3.3.1 飞行时间法································································.50
3.3.2 结构光法···································································.53
3.3.3 莫尔等高条纹法··························································.55
3.3.4 同时采集深度和亮度图像··············································.58
3.4 立体视觉成像······································································.59
3.4.1 双目横向模式·····························································.59
3.4.2 双目会聚横向模式·······················································.63
3.4.3 双目轴向模式·····························································.65
3.5 各节要点和进一步参考··························································.67
第4 章视频图像和运动信息·······························································69
4.1 视频基础············································································.70
4.1.1 视频表达和模型··························································

本书是一本介绍3D 计算机视觉基本原理、典型算法和实用技术的专业图书，建议读者在学习《2D 计算机视觉：原理、算法及应用》后使用本书。
本书在选材上主要覆盖了计算机视觉的进阶级内容，自成体系，主要针对信息类相关专业，同时兼顾了具有不同专业背景的学习者及自学读者的需求。读者可据此开展科研工作并解决实际应用中一些具有一定难度的问题。
本书在编写上比较注重实用性，没有过多强调理论体系，尽量减少公式推导，着重介绍常用的方法。书中有较多的示例，能通过直观的解释帮助读者理解抽象的概念。书末附有术语索引（文中标为黑体），给出了对应的英文，方便读者查阅及搜索相关资料。
本书提供了大量的自我检测题（包括提示和答案）。从目的来说，一方面，这便于自学者判断自己是否掌握了重点内容；另一方面，这便于教师开展网络教学，在授课时加强师生互动。题目类型为选择题，可用计算机方便地判断正误。从内容来看，很多题把基本概念换一种说法进行表达，补充了正文，使学习者能加深理解；有些题列出了一些相似但不相同（甚至含义相反）的描述，通过正反辩证思考，使学习者能深入领会本质。所有自我检测题都附有提示，读者可获得更多的信息以进一步理解题目的含义。同时，在有提示的基础上，如果读者能在看到提示后完成自我检测题，则表明基本掌握了学习内容；如果不看提示就能完成自我检测题，则表明内容掌握得比较好。
本书从结构上看，包括12 章正文、1 个附录及自我检测题、自我检测题答案、参考文献和术语索引。在这17 个一级标题下，共有66 个二级标题（节），再之下有135 个三级标题（小节）。全书共有文字（包括图片、绘图、表格、公式等）50 万字，共有编了号的图228 个、表格22 个、公式565 个。为便于教学和理解，本书给出示例68 个、自我检测题157 道（全部附有提示和答案）。另外，书末列出了直接相关的100 多篇参考文献和用于索引的500 多个术语（中英文对照）。
本书的先修课程知识涉及三个方面。一是数学，包括线性代数和矩阵理论，以及有关统计学、概率论和随机建模的基础知识；二是计算机科学，包括对计算机软件技术的掌握、对计算机结构体系的理解，以及对计算机编程方法的应用；三是电子学，包括电子设备的特性原理及电路设计等内容。
感谢电子工业出版社编辑的精心组稿、认真审阅和细心修改。
后，作者感谢妻子何芸、女儿章荷铭在各方面的理解和支持。
章毓晋
2020 年暑假于书房