基于视觉伺服的机械臂分拣系统研究
发布时间：2018-11-02 15:06
【摘要】：在工业自动化高速发展的背景下,越来越多的机器人技术被运用于工业作业中,这当中就包括机器人分拣技术。传统的分拣技术往往使用运动示教方式,对于目标位置要求必须固定,且有着较低的工作效率。将机器视觉与机械臂相结合,使得机械臂能够自主地识别目标物体,提高了作业速度,降低了劳力成本,有着重大的实际意义。机器人视觉技术使得机器人具备了视觉感知能力,如今也是当今工业自动化的重点研究领域。目前在工业上被广泛用于缺陷检测、物流、码垛、焊接等众多领域。本文以puma560机械臂为研究本体,对于基于位置的视觉伺服控制方法展开研究,设计了一个视觉分拣系统。将目标工件图像处理,工件的匹配识别,相机标定,视觉控制器的设计为主线贯穿全文,最终实现了机械臂对于静止目标和运动目标追踪。本文重点针对以下几点展开了研究:进行了分拣目标图像处理算法的研究和设计,通过实验对比选取了最佳的图像处理方案,去除了目标二值图像的噪声,得到了满意的目标二值图。Otsu算法是图像分割中最常用的方法之一,针对传统Otsu算法需遍历每个灰度值进行类间方差的计算,本文提出一种快速的Otsu改进算法,并与传统算法进行了比较,验证了新算法的有效性。对于分拣目标进行Harris角点特征提取,得到了模版和原图像的角点特征图像。并使用归一化互相关(NCC)策略对原图和模版图进行了粗匹配,得到了包含误匹配的匹配结果。最终通过使用RANSAC策略去除了误匹配,得到了提纯后正确的匹配结果。进行了视觉系统的标定研究,分别对于相机标定和手眼标定方法进行了理论研究与标定实验。计算出了相机的内外参数,并给出了手眼矩阵求解一般方法。对于puma560进行了运动学的求解,研究了基于位置的视觉伺服的控制方法。在matlab/simulink下进行了视觉伺服的控制模型搭建,并分别实现了对静止目标定位以及运动目标追踪的仿真研究。通过上述的研究工作表明,本文提出的视觉系统方案对于目标工件识别定位精度高,对于直线运动目标追踪效果好,证明了本文系统具有一定的实际意义。
[Abstract]:With the rapid development of industrial automation, more and more robot technologies are used in industrial operations, including robot sorting technology. The traditional sorting technology often uses the motion teaching method, which must be fixed for the target position, and has low working efficiency. It is of great practical significance to combine machine vision with robot arm to identify target objects independently, improve the working speed and reduce the labor cost. Robot vision technology makes robot have visual perception ability, and now it is the key research field of industrial automation. At present, it is widely used in many fields such as defect detection, logistics, palletizing, welding and so on. In this paper, the puma560 manipulator is taken as the research body, and a visual sorting system is designed based on the position based visual servo control method. The image processing, matching recognition, camera calibration and the design of vision controller are the main lines in the paper. Finally, the tracking of the stationary and moving targets by the manipulator is realized. This paper focuses on the following points: the research and design of sorting target image processing algorithm is carried out, the optimal image processing scheme is selected through experimental comparison, and the noise of the target binary image is removed. Otsu algorithm is one of the most commonly used methods in image segmentation. In view of the traditional Otsu algorithm needs to traverse each gray value to calculate the inter-class variance, a fast Otsu improved algorithm is proposed in this paper. Compared with the traditional algorithm, the validity of the new algorithm is verified. The corner feature images of template and original image are obtained by Harris corner feature extraction for sorting target. The normalized cross-correlation (NCC) strategy is used to match the original image and template map, and a matching result containing mismatch is obtained. Finally, the mismatch is removed by using RANSAC strategy, and the correct matching result is obtained after purification. The calibration of visual system is studied, and the methods of camera calibration and hand-eye calibration are studied and calibrated. The internal and external parameters of the camera are calculated, and the general method of hand-eye matrix solution is given. The kinematics of puma560 is solved, and the control method of visual servo based on position is studied. The control model of visual servo is built under matlab/simulink, and the simulation research of static target location and moving target tracking is realized respectively. The above research results show that the proposed vision system has high accuracy for target recognition and location, and good tracking effect for linear moving target, which proves that the proposed vision system has certain practical significance.
【学位授予单位】：西安建筑科技大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：TP391.41;TP241
[Abstract]:With the rapid development of industrial automation, more and more robot technologies are used in industrial operations, including robot sorting technology. The traditional sorting technology often uses the motion teaching method, which must be fixed for the target position, and has low working efficiency. It is of great practical significance to combine machine vision with robot arm to identify target objects independently, improve the working speed and reduce the labor cost. Robot vision technology makes robot have visual perception ability, and now it is the key research field of industrial automation. At present, it is widely used in many fields such as defect detection, logistics, palletizing, welding and so on. In this paper, the puma560 manipulator is taken as the research body, and a visual sorting system is designed based on the position based visual servo control method. The image processing, matching recognition, camera calibration and the design of vision controller are the main lines in the paper. Finally, the tracking of the stationary and moving targets by the manipulator is realized. This paper focuses on the following points: the research and design of sorting target image processing algorithm is carried out, the optimal image processing scheme is selected through experimental comparison, and the noise of the target binary image is removed. Otsu algorithm is one of the most commonly used methods in image segmentation. In view of the traditional Otsu algorithm needs to traverse each gray value to calculate the inter-class variance, a fast Otsu improved algorithm is proposed in this paper. Compared with the traditional algorithm, the validity of the new algorithm is verified. The corner feature images of template and original image are obtained by Harris corner feature extraction for sorting target. The normalized cross-correlation (NCC) strategy is used to match the original image and template map, and a matching result containing mismatch is obtained. Finally, the mismatch is removed by using RANSAC strategy, and the correct matching result is obtained after purification. The calibration of visual system is studied, and the methods of camera calibration and hand-eye calibration are studied and calibrated. The internal and external parameters of the camera are calculated, and the general method of hand-eye matrix solution is given. The kinematics of puma560 is solved, and the control method of visual servo based on position is studied. The control model of visual servo is built under matlab/simulink, and the simulation research of static target location and moving target tracking is realized respectively. The above research results show that the proposed vision system has high accuracy for target recognition and location, and good tracking effect for linear moving target, which proves that the proposed vision system has certain practical significance.
【学位授予单位】：西安建筑科技大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：TP391.41;TP241
【参考文献】
相关期刊论文 前10条
1 郑启亮;;复杂背景下配电柜线号定位与识别方法研究[J];计算机测量与控制;2016年10期
2 姚景扬;;机器视觉技术在烟支钢印检测中的应用[J];企业导报;2016年10期
3 倪鹤鹏;刘亚男;张承瑞;王云飞;夏飞虎;邱正师;;基于机器视觉的Delta机器人分拣系统算法[J];机器人;2016年01期
4 尤放;;人机共融开启智能机器人新纪元[J];商业观察;2015年Z1期
5 张薇;于硕;;数字图像处理综述[J];通讯世界;2015年18期
6 王娜;;基于数学形态学的中轴变换算法[J];计算机光盘软件与应用;2014年06期
7 刘军;刘超;;基于CCD视觉的线缆识别技术[J];重庆理工大学学报(自然科学);2014年02期
8 杨莉;潘丰;戴娟;;基于数学形态学和Canny算子的音圈马达磁体边缘检测[J];江南大学学报(自然科学版);2013年06期
9 蔡自兴;郭t，
本文编号：2306145
本文链接：https://www.wllwen.com/kejilunwen/zidonghuakongzhilunwen/2306145.html
