面向人机交互的机器人信息融合系统的研究与实现

发布时间：2018-09-02 05:44

【摘要】：近年来,人工智能取得较大进展,机器人与人的联系越来越紧密,人类和机器人的交互方式种类逐渐增多,但是不管是哪种方式,机器人都是通过传感器来感知人体以及人机交互,这些目标依靠单一的传感器是难以完成的,基于多传感器的信息融合技术是机器人人机交互过程中不可或缺的手段。多传感器的信息融合技术,其实质是对多传感器的信息进行协同优化处理,消除冗余信息,综合各传感器的有效信息形成互补、更加全面的信息,更好的表征机器人的内外环境状态。本文的目标在于研究在人机交互过程中,通过多信息融合的手段,完成机器人对人体目标的跟随功能,使得人机交互更加友好。机器人感知到目标人体之后,在机器人对人体目标的跟随过程中,利用深度信息和颜色信息融合的手段提高机器人视觉跟踪的精度;在后续的人体目标运动跟随过程中,通过基于条件的位姿信息融合的手段,提高机器人运动跟随的精度。本文提出了一种面向人机交互的机器人信息融合系统设计方案,该系统分为三大模块:人体感知模块、机器人视觉跟踪模块、机器人运动跟随模块。本文的工作和创新点如下:1.人体感知模块,我们利用基于运动补偿的帧间差分法检测运动目标,通过骨架验证目标是否为人体目标,以确定机器人后续的动作;2.机器人视觉跟踪模块,研究了小波变换融合算法,我们提出了将Depth图像和反向投影图融合的方式应用在机器人视觉跟踪中的应用,该方案能提高机器人对人体目标质心的视觉跟踪的精度;3.运动跟随模块,在确定人体目标质心运动状态之后,我们提出了机器人视觉位姿和惯性导航位姿信息融合的方式应用在机器人运动跟随中的应用。机器人在运动跟随过程中,实时修正机器人位姿,有效的提高了运动跟随的精度;本文设计和实现了面向人机交互的机器人信息融合系统。并在机器人上实现了该系统和进行相关测试。
[Abstract]:In recent years, artificial intelligence has made great progress, robots are more and more closely connected with human beings, and the types of interaction between human beings and robots are gradually increasing, but no matter which way it is, Robots are aware of human body and human-computer interaction through sensors. It is difficult to achieve these goals by a single sensor. Multi-sensor based information fusion technology is an indispensable means in the process of robot human-computer interaction. The essence of multi-sensor information fusion technology is to optimize the multi-sensor information, eliminate redundant information, synthesize the effective information of each sensor to form complementary and more comprehensive information. A better representation of the robot's internal and external environment state. The purpose of this paper is to study the following function of robot to human body by means of multi-information fusion in the process of human-computer interaction, and make human-computer interaction more friendly. After the robot perceives the target body, it uses depth information and color information fusion method to improve the robot visual tracking accuracy in the process of robot following the human body target. The precision of robot motion following is improved by the method of position and pose information fusion based on condition. In this paper, a design scheme of robot information fusion system for human-computer interaction is proposed. The system is divided into three modules: human perception module, robot visual tracking module, robot motion following module. The work and innovation of this paper are as follows: 1. In the human perception module, we use the motion compensation based inter-frame differential method to detect the moving target, and verify whether the target is a human object through the skeleton, so as to determine the robot's subsequent action. In the robot vision tracking module, the wavelet transform fusion algorithm is studied, and the fusion method of Depth image and reverse projection image is put forward in the robot vision tracking. This scheme can improve the accuracy of robot visual tracking of human target centroid. In the motion following module, after determining the motion state of the human body's center of mass, we put forward the application of robot vision pose and inertial navigation position and pose information fusion in the robot motion following. In the process of robot motion following, the robot position and pose are corrected in real time, which effectively improves the precision of motion following. In this paper, a robot information fusion system oriented to human-computer interaction is designed and implemented. The system is implemented on the robot and related tests are carried out.
【学位授予单位】：电子科技大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：TP391.41;TP242

【参考文献】