摘要
针对工业机器人如何能在多目标工况下快速自主识别和抓取指定目标工件的问题,将单目视觉引导技术应用到工业机器人智能抓取系统设计中。利用图像进行了模式识别,对检测定位进行了研究,建立了视觉图像与工件定位抓取之间的关系,提出了基于轮廓Hu不变矩快速模板匹配算法的单目视觉抓取系统。首先将摄像机获取的图像进行了预处理,然后利用轮廓Hu不变矩模板匹配算法进行了目标工件的识别,利用轮廓矩和二阶惯性矩最小原理对识别出的目标工件进行了位姿求取,最后通过建立SOCKET通信将求取的位姿发送给了机械臂控制系统引导机械臂的抓取。基于VS软件开发平台和ABB机械手,对智能抓取系统进行了搭建并试验。研究结果表明:该基于单目视觉搭建的工业机器人智能抓取系统成本低、定位精度高,可满足工业自动化生产的需求。
Aiming at the problem that an industrial robot how to quickly and autonomously identify and capture a specified target artifact under multi-target conditions,monocular vision was applied to the design of an intelligent grasping system for industrial robots. In order to establish the relationship between the visual image and the workpiece positioning and grapping,the pattern recognition,detection and location using images were studied,and a monocular vision capture system based on contour Hu invariant moments fast template matching algorithm was proposed. Firstly,the image data collected from the camera should be preprocessed,then the target was identified by the template matching method based on the contour Hu invariant moment,and the position of the target workpiece was obtained by using the contour moment and the principle of second-order moment of inertia. Finally,through the establishment of SOCKET communication,the desired pose was sent to the manipulator control system to guide the manipulator to grasp. Based on VS software development platform and ABB robot,the intelligent grasping system was built and test. The results indicate that the industrial robot intelligent grasping system based on monocular vision has low cost,high positioning accuracy and meets the needs of industrial automation production.
出处
《机电工程》
CAS
北大核心
2018年第3期283-287,共5页
Journal of Mechanical & Electrical Engineering
基金
中央高校基本科研业务费专项资金资助项目(2016B02914)
关键词
单目视觉
工业机器人
图像处理
模板匹配
位姿检测
monocular vision
industrial robot
image processing
template matching
pose detection
