摘 要

随着人机交互技术的发展，机器人对用户操作意图的接受程度和理解程度变得越来越人性化。利用人体的简单动作、手势直接控制机器人执行相应命令，可以简化以往复杂繁琐的机器人控制方式，所以基于视觉捕捉技术的人体动作识别已经成为一个重要的研究方向，在人机交互、虚拟现实等领域得到广泛的重视和应用。

作为科技前沿产品，微软公司的Kinect体感传感器系统能够获取彩色图像、深度图像和人体骨骼图像 ，并且为用户提供一种全新的人交互方式。Kinect利用深度图像，完美地解决了二维信息到三维信息重建所引起的重要数据丢失的问题，并且在深度图像的基础上，获取含有三维坐标信息的骨骼数据流。但是由于不同操作者体型的差异和人体动作的错综复杂，很难构建出一套通用的算法来进行人体动作的识别。

本文通过Unity软件，在Unity 3D场景中建立一个人体的3D模型，并将该3D模型的骨骼点与用户骨骼点相关联，结合对Kinect示例程序的理解并在此基础上进程编程，实现了对使用者基本动作的识别效果。

关键词：人机交互；Kinect；动作识别；Unity

Research on Human Motion recognition based on Visual capture Technology

ABSTRACT

With the development of human-computer interaction technology, robot's acceptance and understanding of user's operation intention become more and more humanized. By using the simple action of the human body, the gesture directly controls the robot to execute the corresponding command, which can simplify the complicated and complicated control mode of the robot in the past. Therefore, the research of human motion recognition based on visual capture technology has become an important research direction, and has been widely paid attention to and applied in the fields of human-computer interaction, virtual reality and so on.

As a cutting-edge technology product, Microsoft's Kinect somatosensor system can capture color images, depth images and human skeletal images, and provide users with a new way of interacting with people. Kinect uses depth images. The problem of important data loss caused by the reconstruction of 2D to 3D information is solved perfectly, and the bone data stream containing 3D coordinate information is obtained on the basis of depth image. However, it is difficult to construct a set of common algorithms to recognize human actions because of the difference of body size and the complexity of human actions.

In this paper, a 3D model of human body is set up in the Unity 3D scene by Unity software, and the skeleton points of the 3D model are associated with the user's skeleton point. The understanding of the Kinect example program and the process programming on this basis have been realized, and the recognition effect of the user's basic action is realized.

Key Words：human-computer interaction ; Kinect ; movement identification ; Unity

目 录

第1章 绪论 1

1.1 研究背景及意义 1

1.2 国内外研究现状及趋势 2

1.2.1 人机交互技术的发展 2

1.2.2 人体动作识别 2

1.3 论文的研究内容 4

1.4 论文的结构安排 4

第2章 人体动作识别系统硬件和开发环境介绍 5

2.1 Kinect综述 5

2.1.1 Kinect的发展历史与现状 5

2.1.2 Kinect的结构 5

2.1.3 Kinect的局限 7

2.2 Kinect for Windows SDK 简介及功能介绍 8

2.3 Kinect for Windows SDK应用领域 9

第3章 Kinect结合Unity的软件设计 11

3.1 Kinect 2.0的数据源提取 11

3.2 Unity结合Kinect控制3D模型具体步骤 13

3.2.1 导入Unity资源包插件 13

3.2.2 骨骼点、程序与模型的关联 16

3.3 C#语言编程分析 17

3.3.1 骨骼点空间坐标分析 17

3.3.2 编程处理过程 19

3.3.3 空间坐标转换 20

第4章 Kinect结合Unity的软件调试与循环 22

4.1 Unity 3D模型的调试与显示 22

4.2 Unity软件调试的循环 25

总结与展望 26

致谢 28

参考文献 29

附录 30

附录1.Kinect结合Unity控制3D模型并进行动作识别的总程序 30

附录2.AvatarController.cs程序（即本论文运用的3D模型控制程序） 39