基于目标识别的机械手定位技术研究  

Research on manipulator localization technology based on target recognition

作　　者：刘祚时[1] 穆鸿书 杨国炜 LIU Zuoshi;MU Hongshu;YANG Guowei(School of Mechanical and Electrical Engineering,Jiangxi University of Science and Technology,Ganzhou 341000,China;School of Electrical Engineering and Automation,Jiangxi University of Science and Technology,Ganzhou 341000,China)

机构地区：[1]江西理工大学机电工程学院,江西赣州341000 [2]江西理工大学电气工程与自动化学院,江西赣州341000

出　　处：《传感器与微系统》2021年第12期31-34,39,共5页Transducer and Microsystem Technologies

基　　金：国家自然科学基金资助项目(71361014);江西省研究生创新专项资金资助项目(YC2019—S313)。

摘　　要：采用Eye-to-hand方式固定安装摄像头,实现单目视觉引导的五自由度机械手定位方法。搭建了以STM32F4系列芯片为主控芯片的控制平台,运用蒙特—卡罗数值分析法求解关节机械手工作空间云点图,进行三次多项式插值。最后以三维积木道具作为实验对象,利用改进后的Harris角点检测算法结合人工辅助标识点,在多种规格积木中提取感兴趣目标进行识别和定位。实验结果证明:在世界坐标系参考下Z轴方向的位置平均相对误差为7.096%,Φ偏航角姿态估计平均相对误差为19.58%,在实现目标快速定位上具有一定参考价值。The Eye-to-hand method is used to fix the camera to realize the positioning method of the five-degree-of-freedom manipulator for monocular vision guidance.A control platform with STM32 F4 series chips as the main control chip is built, and the Monte Carlo numerical analysis method is used to solve the cloud point diagram of the joint manipulator working space, and cubic polynomial interpolation is performed.Finally, the three-dimensional building block props are used as the experimental objects, and the improved Harris corner detection algorithm is combined with artificial auxiliary identification points to extract the target of interest from various specifications of building blocks for recognition and positioning.The experimental results prove that the average relative error of the position in the Z-axis direction under the reference of the world coordinate system is 7.096 %,and the average relative error of the Φ yaw angle attitude estimation is 19.58 %,It has certain reference value in achieving the rapid positioning of the goal.

关 键 词：单目视觉 系统标定 轨迹规划 目标定位 

分 类 号：TP391[自动化与计算机技术—计算机应用技术] TH122[自动化与计算机技术—计算机科学与技术]