基于改进Yolov3的H型钢喷号识别系统研究  

Research on H-Beam Spray Number Identification System Based on Improved Yolo v3

作　　者：高磊 徐洪[1] 汪昌平[1] GAO Lei;XU Hong;WANG Changping(Long Product Department of Maanshan Iron and Steel Co.,Ltd.,Maanshan,Anhui 243000,China)

机构地区：[1]马鞍山钢铁股份有限公司长材事业部,安徽马鞍山243000

出　　处：《冶金动力》2024年第3期63-66,74,共5页Metallurgical Power

摘　　要：随着工业生产智能化程度不断提升,智能视频识别技术被逐渐广泛应用于工业生产中。针对马钢重型H型钢生产线精整区存在的工艺复杂、物料跟踪困难问题,提出一种在矫直机后用喷码机在钢坯上喷印批号、在堆垛区域人工检查台位置安装一套字符识别装置来识别批号并实现物料跟踪的方法。根据喷印字符相对较小且没有大小形态变化的特性,改进了Yolov3模型结构,设计5个不同尺度的卷积特征图,并与残差网络中相应尺度的特征图进行融合,形成最终的特征金字塔执行字符识别任务。试验结果表明,该方法的识别准确率能达到99.5%以上,可在工程上进行应用。With the continuous improvement of the intelligent level of industrial produc-tion,intelligent video recognition technology is gradually widely used in industrial produc-tion.Aiming at the complex process and difficult material tracking problems in the finishing area of Masteel's heavy H-beam steel production line,a method of material tracking is pro-posed by spraying the billet with a coding machine after the leveler to print the batch num-ber,and installing a set of character recognition device in the position of the manual check-ing table in the stacking area to recognize the batch number.At the same time,according to the characteristics of relatively small and no size or shape change of the printed charac-ters,the Yolo v3 model structure is improved,five convolutional feature maps with different scales were designed and fused with the corresponding scales in the residual network,so as to form the final feature pyramid to perform the character recognition task.The test results show that the recognition accuracy of this method can reach more than 99.5%and can be applied in practice.

关 键 词：重型H型钢 物料跟踪 字符识别 Yolov3 

分 类 号：TP29[自动化与计算机技术—检测技术与自动化装置]