基于YOLOv4与位置先验的快递三段码检测算法  

Three-segment waybill code detection algorithm based on YOLOv4 and location prior

作　　者：韩贵金 胡仲阳 石海宾 HAN Guijin;HU Zhongyang;SHI Haibin(School of Automation,Xi'an University of Posts and Telecommunications,Xi'an 710121,China;China Construction Second Engineering Bureau Co.,Ltd.North China Company,Beijing 100071,China)

机构地区：[1]西安邮电大学自动化学院,陕西西安710121 [2]中国建筑第二工程局有限公司华北公司,北京100071

出　　处：《西安邮电大学学报》2021年第4期105-110,共6页Journal of Xi’an University of Posts and Telecommunications

基　　金：陕西省教育厅服务地方专项科学研究计划项目(21JC029)。

摘　　要：针对区县一级分拣中心采用的人工分拣方式存在效率低下问题,提出一种基于YOLOv4与位置先验的快递面单三段码检测算法。将检测到的快递面单图像作为快递Logo检测区域,以缩小检测区域,排除与快递面单相似的目标,通过HSV颜色分割对目标区域进行快递面单初检。利用YOLOv4算法对预处理后的快递面单图像进行Logo检测,结合检测到的Logo信息与快递面单中Logo和三段码先验位置关系完成三段码目标检测。实验结果表明,该算法三段码定位准确率达96.1%以上,可有效提高快递分拣效率。For the low efficiency of manual sorting method adopted by the district and county level sorting centers,a three-segment waybill code detection algorithm based on YOLOv4 and location prior is proposed.The detected waybill image is used as the package logo detection zone to narrow the detection area and exclude targets similar to the waybill image.The target area is preliminarily checked by HSV color segmentation.The YOLOv4 algorithm is adopted to perform logo detection for the pre-treated waybill image.Combine the detected logo information with the logo on the package waybill image,and the position relationship to complete the three-segment code target detection.Experimental results show that the accuracy of the three-segment code positioning is over 96.1%,which can effectively improve the efficiency of package sorting.

关 键 词：目标检测 快递三段码检测 YOLOv4 HSV颜色分割 快递分拣 

分 类 号：TP301.6[自动化与计算机技术—计算机系统结构]