基于图像雅可比矩阵的无标定二维运动跟踪  

UNCALIBRATED 2D MOTION TRACKING BASED ON IMAGE JACOBIAN MATRIX

作　　者：钱江[1] 苏剑波[1] 

机构地区：[1]上海交通大学自动化研究所智能机器人系统与技术研究中心,上海200030

出　　处：《模式识别与人工智能》2003年第3期257-262,共6页Pattern Recognition and Artificial Intelligence

摘　　要：研究无标定视觉导引机械臂完成二维平面运动目标精确跟踪的问题。系统的视觉部分采用固定全局视觉与手上局部视觉相结合的方式,拓展了系统的观察能力,使系统能同时实现大范围的运动跟踪和精确的小范围跟踪。根据两种视觉的特点,在全局视觉反馈中使用最优反馈控制,而在手上视觉反馈中使用PI控制。通过分析图像雅可比矩阵的解析表达式以及跟踪过程中的变化规律,提出一种利用机器人已完成的跟踪运动在线辨识雅可比矩阵的方法,避免了传统估计方法引入冗余试探运动对系统动态性能的影响。仿真和实验检验了本方法的性能。This paper presents a visual feedback control scheme for the task of accurate 2D motion tracking in the uncalibrated environment. A hybrid eye configuration with a global fixed camera and an eye in hand is used to extend the observing ability and help the system to track moving target in large-scale. According to the difference between two cameras, the optimal control rate is used in global vision feedback and the PI controller is used in eye-in-hand visual feedback. The variation of the Image Jacobian matrix during tracking motion was discussed by its analytic expression. We also propose an online estimation algorithm for image Jacobian without redundant try movements. Simulations and experiments show the tracking accuracy of this scheme.

关 键 词：无标定机器人 手眼系统 智能机器人 机械臂 图像雅可比矩阵 无标定二维运动跟踪 

分 类 号：TP24[自动化与计算机技术—检测技术与自动化装置]