攀爬机器人末端位姿的视觉检测与控制  

Visual Detection and Control for Gripper Posture of Climbing Robot

作　　者：吴俊君[1,2] 胡国生[1] 

机构地区：[1]广东食品药品职业学院,广州510520 [2]华南理工大学机械与汽车工程学院,广州510640

出　　处：《计算机测量与控制》2013年第7期1740-1742,共3页Computer Measurement &Control

基　　金：国家自然科学基金资助项目(50975089);高等学校博士学科点专项科研基金资助项目(200805611091)

摘　　要：攀爬机器人末端夹持器位姿的自主调整是实现机器人自主攀爬的重要环节;在利用激光或声纳传感器进行位姿倾斜度检测时,会出现"夹持器摇摆现象",检测精确度不高而且耗时;针对这个问题,提出一种用于夹持器位姿倾斜检测的视觉检测器;检测器会在基于各向异性扩散滤波构建的多尺度空间提取图像的边缘信息,然后识别属于杆件的边缘直线,完成位姿倾斜度检测与控制;实验结果显示:检测器能较好地识别杆件边缘直线,精确计算出夹持器相对被夹杆件的倾斜度,时间约为0.8s;该检测器简洁实用,能较好地满足自主攀爬控制系统对精度和速度的需求。Self-- control of gripper position and orientation for climbing robot is an important part of biped climbing. Laser or sonar detec- tion would lead to the phenomenon of gripper swing, which has low accuracy and real time. For the problem, a visual posture detector for gripper of climbing robot has been proposed. Firstly, multi-scale image spatial was constructed by anisotropic diffusion filtering the edges in image were extracted, the max edge belong to pole was recognized, and then the posture of gripper was computed. Experimental results in- dicated that the detector recognized edge of pole well, the inclination of gripper was accuracy, and the cost time is about 0.8 s. The detector is simple and practical, which worked well for climbing control system.

关 键 词：仿生攀爬机器人 自主攀爬 夹持器位姿 视觉检测器 

分 类 号：TP242[自动化与计算机技术—检测技术与自动化装置]