《单幅图像单目视觉测距方法适用性分析》由会员分享，可在线阅读，更多相关《单幅图像单目视觉测距方法适用性分析（56页珍藏版）》请在金锄头文库上搜索。
1、 硕士学位论文硕士学位论文 单幅图像单目视觉测距方法适用性分析 APPLICATION ANALYSIS OF DEPTH EXTRACTION FROM SINGLE MONOCULAR IMAGE 赵 信 宇 哈尔滨工业大学哈尔滨工业大学 2009 年 12 月 国内图书分类号：TP242 学校代码：10213 国际图书分类号：620 密级：公开 工工学硕士学位论文学硕士学位论文 单幅图像单目视觉测距方法适用性分析 硕 士 研 究 生 ： 赵信宇 导 师 ： 朱晓蕊 副教授 申请学位 ： 工学硕士 学科 ： 控制科学与工程 所 在 单 位 ： 深圳研究生院 答 辩 日 期 ： 2009年
2、12 月 授予学位单位 ： 哈尔滨工业大学 Classified Index: TP242 U.D.C: 620 Dissertation for the Master Degree of Engineering APPLICATION ANALYSIS OF DEPTH EXTRACTION FROM SINGLE MONOCULAR IMAGE Candidate: Xinyu Zhao Supervisor: Prof. Xiaorui Zhu Academic Degree Applied for: Master of Engineering Specialty: Control Sc
3、ience and Engineering Affiliation: Shenzhen Graduate School Date of Defence: December, 2009 Degree-Conferring-Institution: Harbin Institute of Technology -I- 摘 要 距离信息在物体识别、路径规划以及场景恢复上都有着重要应用。当我们行走的时候，我们可以轻易地判断出障碍物距离我们有多远，但是对于笨拙机器人来说，这个任务就变得相当艰巨了。随着机器人导航在机器人应用里变得越来越重要，距离检测也就成了大家研究的重点。在机器人和机器视觉领域，距离估计
4、可以通过摄像头或者其他图像处理的方法，一般常用的是立体视觉法。但是立体视觉会受限与两个摄像头基线之间距离，还会受到亮度噪声的影响，当然，利用两个摄像头会花费更多的成本。因此，我们考虑用一个摄像头来完成距离估计的任务。 本文首先介绍了一些有关机器视觉和高斯-马尔科夫随机理论的基础知识，这些知识将在我们随后的建模中用到。然后我们依据人类的视觉系统，建立起概率模型。为了训练该模型，我们需要采集一些图片以及他们的相关深度图，为此我们设计了一套激光测距系统。最后，通过训练参数完善模型，并将其应用于新的图像。结果表明，此方法能够实现图像深度信息的提取，并可以应用于机器人导航。 关键词：距离估计；高斯-马尔
5、科夫理论；单目计算机视觉；激光测距系统 -II- ABSTRACT In the researches of object recognition, path planning and scene interpretation, depth information has an important use. Humans can easily estimate how far away from the obstacles when we walk, but for clumsy robots, it becomes too difficult to judge how far away ob
6、jects are. Along with the robotic navigation technology has become the focus of the mobile robot industry, how to estimate the distance has become an important task. Depth estimation has variant methods by using camera and image processing, which in robotics and computer vision mostly uses stereo vi
7、sion. Stereo vision is fundamentally limited by the baseline distance between the two cameras, and it may suffer from sudden changes of image intensity due to the illumination noise, also using two cameras makes the cost much expensively. Thus we consider the task of depth estimation from a single m
8、onocular image. We first introduced some basic theories of machine vision and Gaussian-Markov theorem which we will use to build model. On the basis of humans vision system, we build the probabilistic model. To collect images and their depthmaps, we designed the laser scanner system. At last, we tra
9、ined the parameters of the model and test for new images. The result proved that this method can estimate depth information from monocular images, and can be used for robot navigation. Keywords: depth estimation, Gaussian-Markov theorem, monocular computer vision, laser scanner system -III- ACKNOWLE
10、DGEMENTS My deepest gratitude goes first and foremost to Professor Xiaorui Zhu, my supervisor, for her constant encouragement and guidance. She has walked me through all the thesis. Without her invaluable help and generous encouragement, the present thesis would not have been accomplished. Secondly,
11、 I would like to express my heartfelt gratitude to all the teachers, from whose devoted teaching and enlightening lectures, I have gained profound knowledge, remarkable expertise and intellectual ingenuity, all of these will give much help to my future life and career. Besides, I would like to thank
12、 my colleagues Tao Zhao, Chunxin Qiu, Yanmin Zhang, Shuyan Cui, and Jinquan Yu, who helped me search for reference. I also owe my sincere gratitude to my friends and my fellow classmates who gave their help and time in listening to me and helping me work out my problems during the difficult course o
13、f the thesis. Last I would like to express my gratitude to my beloved parents who have always been helping me out of difficulties and supporting without a word of complaint. My heart swells with gratitude to all the people who helped me. -IV- CONTENTS 摘摘 要要 . I ABSTRACT . II ACKNOWLEDGEMENTS .III LIST OF FIGURES . VI CHAPTER 1 INTRODUCTION .1 1.1 Motivation .1 1.2 Current research .2 1.3 Organization of the Dissertation .
