基于深度摄像机的航天员手势识别方法研究

发布时间：2018-02-26 12:11

本文关键词： 空间机器人控制技术体感技术人机交互手势识别图像处理　出处：《哈尔滨工业大学》2017年硕士论文　论文类型：学位论文

【摘要】：依托空间机器人控制技术的高速发展,宇航员和空间机器人的交互活动愈发频繁和普遍。机器视觉领域中的手势识别方法研究对于实现新型人机交互方式将起到极为关键的作用。本论文的研究目的是期望建立一种航天员通过手势操纵空间机器人或飞行器全向运动并辅助完成空间任务的新型控制系统,这对于我国未来空间应用技术的发展有很高的研究价值和发展前景。本论文的主要研究内容是基于深感摄像机实现了人体静态手势和动态手势的动作识别。此外,对文中的手势识别控制系统在空间站服务型机器人上的应用做了可行性分析。最后,设计了一套操作者通过手势动作控制轮式机器人全向运动的硬件系统,从而完成了这种新型无接触式的手势控制方式的方法验证。本论文中静态手势识别的研究是使用本文提出的基于肤色检测和深度阈值相结合的方法进行图像中手部区域的分割和提取。然后,通过对图像降噪处理和细化处理获得手部骨架图并通过文中提出的基于手骨架图的特征点提取方法进行手部关键节点的识别和提取。最后,通过建立静态手势三维模型实现了具有更高准确性和快速性的静态手势识别。文中动态手势识别的研究是基于改进的动态时间规整方法通过将采集到的手势动作运动轨迹序列与预先建立的标准动作模板库进行对比和匹配。实验测试结果表明,文中改进后的动态时间规整方法实现了具有更高准确性和快速性的动态手势识别。此外,本文把手势控制应用到机器人控制系统中,实现了一种新型的机器人控制方式。同时,对于宇航员手势控制空间站服务型机器人做了可行性和可用性的仿真分析与验证。
[Abstract]:Relying on the rapid development of space robot control technology, The interaction between astronauts and space robots is becoming more and more frequent and common. The research of gesture recognition in the field of machine vision will play a crucial role in the realization of a new type of human-computer interaction. The purpose of this paper is to. Desiring to establish a new type of control system in which astronauts use gestures to manipulate the omnidirectional motion of space robots or aircraft and to assist in the completion of space missions, This will be of great value and prospect for the future development of space application technology in our country. The main research content of this thesis is to realize the motion recognition of human body static and dynamic gestures based on deep sense camera. The feasibility of the application of the gesture recognition control system in the space station service robot is analyzed. Finally, a hardware system is designed to control the omnidirectional motion of the wheeled robot by hand gesture. In this paper, the method of static gesture recognition is based on the combination of skin color detection and depth threshold. The segmentation and extraction of the region. Then, The hand skeleton map is obtained by image denoising and thinning processing, and the hand key nodes are identified and extracted by the method of feature point extraction based on hand skeleton image proposed in this paper. The static gesture recognition with higher accuracy and rapidity is realized by establishing 3D static gesture model. The research of dynamic gesture recognition in this paper is based on the improved dynamic time regularization method. The motion trajectory sequence is compared and matched with the pre-established standard action template library. The experimental results show that, The improved dynamic time warping method realizes the dynamic gesture recognition with higher accuracy and rapidity. In addition, this paper applies the gesture control to the robot control system to realize a new robot control mode. Simulation analysis and validation of the feasibility and availability of the space station service robot with astronaut gesture control are carried out.
【学位授予单位】：哈尔滨工业大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：TP242;V441

【参考文献】