Kinect基于手势识别的三维场景视点控制算法研究
毕业论文关键词：SVM Hu 不变矩 静态手势识别 动态手势识别 三维场景控制
Title Study on viewpoint control algorithm of 3D scenes by gesture recognition
Abstract Gesture recognition algorithms based on skeleton and depth information captured by Microsoft Kinect are studied in the thesis and a real-time and online gesture recognition system is designed. Furthermore, the gesture recognition system is integrated into a 3D scene viewpoint control system to manipulate scene changes through gesture. Both static and dynamic gesture recognition methods are discussed in the thesis. As for static gesture recognition, an algorithm consists of three steps is proposed that roughly locates hands based on depth data and skeleton points of Kinect, precisely segments out hands through horizontal and vertical projections and accomplishes gesture recognition using Hu invariant moments and SVM. While the dynamic gesture recognition is to obtain kinetic features via the movement trajectories of hands extracted from the depth data streams and post-processed with uniformly-spaced sub-sampling. Three dynamic gesture recognition algorithms are explored and finally a method combining the template matching of moving trajectories and the SVM classification of the tangent angles of the trajectories is adopted. In practical usage, the proposed method can achieve a recognition rate over 90%. Finally, a 3D-scene viewpoint control system is designed and implemented that uses the dynamic gesture recognition algorithm to manipulate a 3D scene viewer realized by OSG. Keywords: SVM, Hu invariant moments, Static gesture recognition,
Dynamic gesture recognition，3D scene control
目录
1绪论···1
1.1选题背景和意义··1
1.2手势识别研究现状简介·2
1.3论文完成工作及章节安排··3
2手势识别算法研究·4
2.1Kinect简介4
2.2静态手势识别·5
2.3动态手势识别12
3基于手势识别的三维场景控制系统实现·20
3.1系统结构框图20
3.2静态手势识别的实现···21
3.3动态手势识别的实现···24
3.4三维场景漫游实现··26
3.5Android手势识别实现·29
4系统运行结果··33
4.1实验系统运行结果··33
4.2基于手势识别的三维场景控制系统运行结果35
5总结与展望·36
5.1总结36
5.2展望36
致谢39
参考文献40
上一篇：火蔓延LED点阵屏显示模拟系统
下一篇：ARM7永磁开关控制器软件设计