【作者】 杜荣；

【导师】 焦恩璋；

【作者基本信息】 南京林业大学 ， 机械设计及理论， 2009， 硕士

【摘要】 本文以Motoman up6机器人为基础,研究并构建了一个基于机器视觉的工业机器人分拣系统。该系统由相机、图像采集卡、计算机、机器人以及软件组成。该系统通过相机获取作业对象图像,然后由软件对采集到的图像进行分析,得出目标的坐标与分类信息并维护目标的运动踪迹,最终控制机器人实现分拣动作。本文首先研究了相机标定技术,在Matlab上实现了直接线性算法,并提出了一个将世界坐标系转换到机器人坐标系的简易方法,。然后,研究了多种目标检测算法原理,着重研究并实现了基于背景减法的目标检测以及基于灰度图二值化的目标检测。在目标识别方面,本文根据作业对象的面积、周长等特征,采用线性判别函数法设计并训练分类器。在视觉跟踪方面,研究了全局最近邻法与多假设跟踪法,通过编程实现了GNN算法与卡尔曼滤波,实现了传送带上的目标跟踪。最后,本文在Motocom32的基础上研究了基于PC机的Motoman机器人运动控制,使得分拣程序可以控制机器人在传送带连续运动时实时抓取带上的工件。在该研究过程中还发现了Motocom32的一些不足,并提出了解决方法。经过多次实验和调试,作者完成了一个初级的工业机器人分拣系统的研究和构建。联机实验证明了本文的研究成果,该系统作为工业自动化的一个环节,在应用上有着较好的实用意义,在理论上也有一定的参考价值。更多还原

【Abstract】 This paper based on Motoman up6 robot researches and structures a classification-and-grab system on the basis of machine vision. The system is made up of camera, image acquisition card, computer, robot and software. The system controls the robot to achieve picking and choosing in the way that the robot getting image of the object by the camera, analyzing the image with software, getting the coordinate and classification data of the object, then track the targets.At first, the paper studies the camera calibration, then implements the Direct Linear Transform algorithm on Matlab, and provides an easy solution to transform word coordinates into robot coordinates. Then, the paper studies several target detection algorithm principles, concentrates on and implements the target detection based on background subtraction and gray image thresholding. At the aspect of target recognition, this paper design and train the classifier by using the linear discriminant function according to the object’s feature such as area and perimeter. At the aspect of vision track, this paper studies global nearest neighbor algorithm and multi-hypothesis tracking algorithm, implements the GNN algorithm and kalman filter in the c++ programming language and finally achieves the target track on conveyor. Finally, the paper researches the PC-based motion control of Motoman robot on the basis of Motocon32, so that the classification-and-grab programme can control robots grab the workpiece on contiguous moving conveyor in real time. In the process of research, some limitation of Motocom32 has been found, and the solution is proposed also.Through several experiments and tests, the author accomplishes the research and structure of a primary industrial robot classification-and-grab system. On-line experiments prove the research results of this paper that is this system as a link of industrial automation has practical significance in applications and reference value in theory.更多还原