【作者】 田祥庆；

【导师】 闵华清；

【作者基本信息】 华南理工大学 ， 计算机应用技术， 2010， 硕士

【摘要】 本文的研究紧紧依托于华南理工大学智能机器人实验室的家政机器人平台,针对家政机器人进行了详细的分析,获得了一些较重要的研究成果和结论。首先构建了包括摄像机标定、目标物图像的获取和识别、获取目标点的深度信息、空间坐标变换、机器人运动学模型分析与建立、以及基于位置的视觉伺服在内的双目机器人视觉伺服控制系统的框架结构。在理论上建立了基于D-H表示法的家政机器人手臂的运动学模型,对其进行了逆运动学求解,并探讨了动态的机器人手眼协调关系。针对SIFT目标识别算法的特点及其在机器人目标识别应用中的不足,提出了结合Snake模型的SIFT目标位置及轮廓识别算法,解决了家政机器人目标识别过程中目标物轮廓信息获取不足的问题,并将相关的目标信息结合到机器人手臂末端执行器对目标物的抓取过程中。利用Visual Studio 2005开发平台实现了相关算法。研究了机器人的视觉伺服系统,对传统的基于位置的视觉伺服系统的结构及相关技术进行了分析,根据其在机器人目标物抓取中的不足,结合雅可比矩阵在机器人微分运动学中的关系,提出了改进的基于位置的视觉伺服系统控制结构。该控制结构的主要思想是通过运动学逆解使家政机器人手臂达到目标物的大致位置,接下来在线计算机器人的雅克比矩阵以控制手臂的微分运动,最后使其能够精确的达到目标物位置并抓取目标物。并通过实验证明了本文中以上理论的有效性。更多还原

【Abstract】 This research closely relies on the household robot platform in the intelligent robot Lab of South China University of Technology, analyzing the household robot particularly. This study gets access to some important research results and conclusions.First it builds binocular robot vision servo control system’s framework, including camera calibration, target image acquisition and processing, getting the depth information of the target point, space coordinate transformation, kinematics modeling and analysis, and visual servo control system. Then we establish the kinematics model for our household robot based on the D-H method, and the inverse kinematics was carried out. We research the relationship of the dynamic robot hand-eye coordinate transformation.This study particularly analyses the characteristics of SIFT object recognition algorithm and its application in robot target recognition, we propose a method which combines Snake model and SIFT target recognition algorithm to solve the issue that we can’t obtain enough target contour information from the traditional robot target recognition process, and it integrate the target information to the target capture process using robot arm end-effector. We use Visual Studio 2005 development platform to achieve the correlation algorithm.This paper researches the robot visual servo system, and analyses the related techniques of the traditional position-based visual servo system. According to the traditional method’s deficiency in the target capture process with robot hand, and the relationship of the Jacobian matrix in robot differential kinematics, we propose the improved position-based visual servo system control method. The main idea of this control method is to control the robot hand to reach the target approximate location using inverse kinematics, then to calculate the Jacobian matrix of the robot hand on line, and finally to enable robot hand to reach the target precise position and grasp the target object. We Proved the effectiveness of this theory by experiment.更多还原