直接零件标志条码区域定位算法  

Direct Part Mark Barcode Location Method

作　　者：王娟[1] 王萍[1] 

机构地区：[1]天津大学电气与自动化工程学院模式识别实验室,天津300072

出　　处：《计算机辅助设计与图形学学报》2014年第7期1159-1166,共8页Journal of Computer-Aided Design & Computer Graphics

摘　　要：针对受工况、光照的限制,现场采集的含有直接零件标志(DPM)条码的图像经常亮度不一,并伴有变形等情况,导致因定位不准确而识别DPM条码困难,提出一种改进的SUSAN角点检测和半监督机制下的近邻传播聚类相结合的DPM条码定位算法.首先针对DPM条码图像区域多直角的特点对检测角点的SUSAN算法进行改进,使其适应由于光照问题导致灰度不均的DPM条码区域,具有检出变形角点的能力,同时极大程度地抑制背景上及条码区域中的伪角点;以此为基础,使用近邻传播聚类算法对角点进行聚类,并将DPM条码区域中角点的聚集特点转化为监督规则,形成半监督机制下的角点近邻传播聚类.实验结果表明,该算法检测出的角点对DPM条码区域的示意性强、效率高,基于角点聚类结果的DPM条码定位比其他算法精准、快速.Affected by working condition or uneven distribution, Direct Part Mark（DPM） barcode image taken by camera is always brightness uneven and distortion. It is difficult to locate and identify DPM. Accordingly, a method based on improved SUSAN corner detection and Semi-supervised Affinity Propagation Clustering is proposed in this paper. After analyzing the characteristics of rectangular areas on DPM barcode image, SUSAN corner detection algorithm is improved in this paper. The improved algorithm has the ability of solving the uneven lighting problems and detecting the distortion barcode. At the same time, it can significantly inhibit off the false corner on the background and within barcode region. Then the rule of supervision on the basis of gathering characteristics of corner is built up. Supervision mechanism is integrated into the Affinity Propagation Clustering to make up Semi-supervised Affinity Propagation Clustering algorithm. Experiments show that our algorithm is high efficiency and strong in detecting the corner within DPM barcode region. DPM barcode location method based on clustering results is more accurate and faster than other algorithm.

关 键 词：DPM条码 角点检测 半监督近邻传播聚类算法 条码定位 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]