CASIA OpenIR  > 毕业生  > 博士学位论文
数字娱乐中的人体特征提取与运动分析研究
其他题名Human Features Extraction and Motion Analysis in Digital Entertainment
曾祥永
学位类型工学博士
导师王阳生
2007-06-20
学位授予单位中国科学院研究生院
学位授予地点中国科学院自动化研究所
学位专业模式识别与智能系统
关键词计算机视觉 特征提取 人脸重建 拳头检测 背景分割 运动分析 人机交互 数字娱乐 Computer Vision Features Extraction Face Reconstruction Fists Detection Background Segment Motion Analysis Computer Human Interaction Digital Entertainment
摘要本文针对数字娱乐用户实际需求及其应用环境,围绕人机交互主体人的身体的局部和整体,研究了基于计算机视觉的快速人体特征提取与人体运动分析方法和技术,并应用这些技术与方案开发了典型的数字娱乐应用,对将视觉交互技术研究与数字娱乐应用开发相结合进行了有益的探索。 本文的主要工作和贡献包括: 1. 提出了一种基于脸部特征点检测的个性化人脸建模的快速方法。为建立个性化的人脸模型,文中首先讨论了基于单目视觉的脸部正面特征点的检测方法。在获取正脸特征点之后,我们采用径向基函数RBF 方法对标准模型进行变形并重新生成纹理贴图,自动生成个性化3D 人脸模型。我们提出了该方法在网络游戏中的应用技术和方案,并设计实现了网络游戏角色置入系统。 2. 提出了贝叶斯框架下的基于颜色和运动及外观信息的手部检测和跟踪方法。为实现应用双手的自然交互,我们用人脸位置的颜色信息建立肤色模型,并根据图像差分方法获取运动信息。结合肤色模型和运动信息以及表观信息,我们实现了对手部位置的精确检测与跟踪。我们提出了该方法在模拟驾驶中的应用方案,并设计开发了利用双手运动模拟汽车和飞行驾驶操作的互动娱乐应用。 3. 提出了基于时间差分和背景减除相结合的快速背景分割与人体提取方法。为实现更为普遍的交互手段,我们将人体作为一个整体的控制设备。联合应用时间差分和背景减除,我们实现了快速的背景分割和人体提取,进一步应用运动历史图像HMI 方法进行运动分析,提取人体运动参数和识别简单的人体姿势。应用这些技术,我们设计开发两类应用系统:面向家庭环境个人用户的、基于普通USB 摄像头的自然互动游戏;面向公共场所群体用户的、允许参与者与投影图像实时交互的互动投影系统。 本文工作以研究和应用快速、鲁棒及低成本的视觉交互技术为目标,各项实验和应用表明,本文方法和技术方案具备较高的可行性和实用性。
其他摘要This dissertation revolves about the main part of HCI ---human's partial and whole body, in view of the applied environment and its user's actual demand of digital entertainment, researches deeply the theory, fast approach and robust technology of computer vision based human feature extraction and motion anal- ysis, and then applies and the results to develop typical digital entertainment applications. And exploit the approach to syncretize vision based interactive technology and digital entertainment. The main contributions of this dissertation are as follows: 1. In order to generate personalized 3D face model, facial features detection and tracking are firstly discussed in this dissertation. we presented a fast personalized face model generation approach based on face features and developed the applied system. With a scheme of data transmission and 3D generation, this system can generate personalized 3D face model for multiuser in networked applications in real time. 2. For realize natural HCI using two hands of human, the motion and colour based approach to detect and track two fits is discussed in this dissertation. We used the color of face region in image to generate skin color model, and apply image time difference to get motion information. By combining color and motion information, two fists can be detected and tracked accurately with high running speed. Based on the proposed fists tracking technique, we designed and developed a driving and a flying simulative game applications. 3. In order to achieve more common interaction, human body can be used as a whole interaction tool. So, image time difference based human motion detection and background segment based human body extraction and MHI(Motion History Image) based motion analysis techniques are also discussed in this dissertation. Applying these techniques, we developed several natural interactive games using one single common PC camera as input device. At the same time, we also developed a new kind of interactive media based on infrared image ---allowing the participants to affect the projected image by the motion and gesture of their bodies. In this dissertation, our goal is to researched and develop fast, robust and low cost approaches of computer vision based partial and whole human body features extraction and motion analysis, which adapt to entertainment applications. The presented tests and typical applications prove that these techniques and ideas have high feasibility and practicality.
馆藏号XWLW1142
其他标识符200318014603042
语种中文
文献类型学位论文
条目标识符http://ir.ia.ac.cn/handle/173211/6032
专题毕业生_博士学位论文
推荐引用方式
GB/T 7714
曾祥永. 数字娱乐中的人体特征提取与运动分析研究[D]. 中国科学院自动化研究所. 中国科学院研究生院,2007.
条目包含的文件
条目无相关文件。
个性服务
推荐该条目
保存到收藏夹
查看访问统计
导出为Endnote文件
谷歌学术
谷歌学术中相似的文章
[曾祥永]的文章
百度学术
百度学术中相似的文章
[曾祥永]的文章
必应学术
必应学术中相似的文章
[曾祥永]的文章
相关权益政策
暂无数据
收藏/分享
所有评论 (0)
暂无评论
 

除非特别说明,本系统中所有内容都受版权保护,并保留所有权利。