成都电子科技大学2013届本科毕业论文.doc

摘要 人体姿态估计是计算机视觉领域中一个关键问题，可以应用于人体活动分析，人机交互以及视频监视等方面主要是指从图像中检测出人体各位置方向尺度信息。人体姿态估计常常被人们在视频跟踪环境中提起近年来人体姿态估计开始转到静态图片人体姿态估计。本文正是从计算机视觉出发，对于基于视觉的静态图片中人体姿态估计进行了研究，主要做了以下工作： 了Kinect技术的一些原理和在人体姿态估计当中的应用；了一些基本的图像特征提取深度信息，梯度直方图（）和形状上下文（Shape Context）的算法原理及应用现状；最后，针对不同的三种图像特征分别进行实验，从得到的实验结果中，平均绝对误差的大小分析比较了三种不同特征下的人体姿态估计之间的。 ABSTRACT Human pose estimation is an essential issue in computer vision area since it has many applications such as human activity analysis, human computer interaction and visual video surveillance, it main purpose of human pose estimation is that detect the position、scale and direction of parts of people .Human pose estimation is often approached in a video setting, within the context of tracking. Recent focus in the area has expanded to single-image pose estimation, because of its’ foundation and convenience . In this dissertation, vision-based human body estimation is investigated. Main contributions of this thesis are follows： e get to know some main principals about the technology of the Kinect，and its real application in the human pose estimation.e do some research about the extract of picture features，such as depth informationhistogram of gradient（）and shape context（Shape Context）.In particular，we explain the main principals of all these methods and their statement of applications.We do some research about the principal of PCA and we use it to reduce our datas’ dimensions; At last，we perform a lot of experiments with respect to the picture features mentioned above ，judging from the results obtained from experiments，we analyse these three different experiments’absolute average errors used in human pose estimation. Keywords：human body estimationKinect,depth information,histogram of gradient ，shape context,PCA 目 录 第1章 引言 1 1.1 绪论 1 1.2 人体姿态估计的研究意义 1 1.3 人体姿态估计研究现状 3 1.3.1 人体姿态估计分类 3 1.3.2 静态图片中的人体姿态估计 3 1.4 人体姿态估计研究难点 4 1.5 本文的研究内容和结构安排 5 1.5.1 本文的研究内容 5 1.5.2 本文的结构安排 5 第2章 图像深度信息 7 2.1 深度图像的研究现状 7 2.1.1 深度图像的概念与特征 7 2.1.2 深度图像研究现状 8 2.2 Kinect技术 8 2.2.