面向工业机器人的平面工件识别与定位方法研究

CORC > 沈阳自动化研究所 > 中国科学院沈阳自动化研究所 > 光电信息技术研究室

题名	面向工业机器人的平面工件识别与定位方法研究
作者	唐宇
学位类别	硕士
答辩日期	2015-05-26
授予单位	中国科学院沈阳自动化研究所
授予地点	中国科学院沈阳自动化研究所
导师	吴清潇
关键词	工件分拣形状特征目标识别模板匹配工件定位
其他题名	Research of Plane Parts Recognition and Positioning for Industrial Robots
学位专业	控制工程
中文摘要	工件分拣是工业机器人的典型应用之一，其中工件的识别与定位是完成分拣操作的必要条件。工件特点不同，视觉识别和定位的方法也不同，其中对平面的具有直线外轮廓的工件在有重叠和粘连的情况下的分拣是典型的应用之一。本文在这一约束条件下，开展了工件的识别与定位方法研究，具有一定的理论意义与很好的应用价值。本文的主要工作和研究内容如下：（1）直线段基元提取方法：本文介绍了基于边缘点的直线段提取算法和基于相位编组的直线段提取算法，并对比分析了这两类算法中的三个代表方法，经典的Hough变换、EDLines和LSD。对比这三种方法在场景图像中的提取结果，选择LSD作为本文的直线段提取算法。（2）基于区域划分的平面工件识别定位方法：在LSD所提取直线段的基础上，先对所提取的直线段进行杂线段的剔除和直线段的连接，然后采用右手摸墙算法将由直线段划分的各个独立的区域提取出来，并形成各区域的形状特征向量，最后与模板工件的形状特征向量进行匹配，完成目标工件的识别和定位。（3）实验：对比分析了所提算法和几何哈希算法在工件粘连和工件重叠两种情况下的实验结果，结果验证了本文所提算法的有效性和鲁棒性。
索取号	TP242/T26/2015
英文摘要	Industrial parts sorting is one of the typical applications of industrial robots, recognition and positioning of the industrial parts is a necessary condition for the completion of the sorting operation. According to different characteristics of different industrial parts, the methods of visual identification and location are different, wherein the sorting of plane industrial parts which have outer contour of straight lines under the condition of overlap and adhesion is a typical case. Under this constraint, this paper carried out studies of recognition and positioning of plane industrial parts, which has some theoretical significance and practical value. The main work and research contents are as follows: Line segment extraction method: This paper describes the line segment extraction algorithms based on edge point and line segment extraction algorithms based on phase classification. Among these two types of algorithms, this paper comparatively analyzes three classic algorithms, classical Hough transform, EDLines and LSD. Compared the results of these three methods on the scene image, this paper selects LSD as the line segment extraction algorithm. Recognition and positioning algorithm of plane industrial parts based on regional division: On the basis of line segments extracted with LSD, firstly the algorithm performs line segments connecting and miscellaneous line segments removal, then uses wall-following algorithm to extract each region divided by line segments and form the shape feature vector of each region, the finally matches shape feature vector with template to complete the recognition of positioning of industrial parts. Experiment: A comparative analysis of the proposed algorithm and Geometric Hashing under overlapping and adhesion circumstances is presented, the results show the effectiveness and robustness of the proposed algorithm.
语种	中文
产权排序	1
页码	61页
内容类型	学位论文
源URL	[http://ir.sia.ac.cn/handle/173321/16777]
专题	沈阳自动化研究所_光电信息技术研究室
推荐引用方式 GB/T 7714	唐宇. 面向工业机器人的平面工件识别与定位方法研究[D]. 中国科学院沈阳自动化研究所. 中国科学院沈阳自动化研究所. 2015.