成捆特钢棒材端面字符识别算法研究

张付祥 郭旺 黄永建 王春梅 黄风山

摘 要：為实现特钢棒材生产信息的全流程可追溯，分析特钢棒材生产环境及形状特点，采用基于双标志点的标记方案进行标记并利用机器视觉技术实现成捆特钢端面字符的识别。首先，采用Hough变换将成捆特钢棒材端面图像分割成单根;其次，采用基于小波变换的图像增强算法完成单根特钢棒材端面图像的增强;再次，将MSER算法和边缘检测算法相结合完成单根特钢棒材字符区域的检测，并基于投影法完成字符的分割;最后，通过创建和训练SVM分类器完成每一根特钢棒材端面字符识别，并将成捆特钢棒材端面字符识别结果输出保存。结果表明，新算法可以满足成捆特钢棒材生产过程中字符识别要求，字符识别准确率可达到97.35%。新算法将Hough变换、基于小波变换的图像增强算法、MSER算法、边缘检测算法、投影法及SVM分类器等算法融合到特钢棒材端面字符识别过程中，为特钢棒材生产过程中的信息获取、信息传递及信息追溯的技术实现提供了参考。

关键词：计算机图像处理;特钢棒材;信息标记;图像增强;字符分割;字符识别

中图分类号：TP391.43   文献标识码：A

doi：10.7535/hbkd.2021yx05005

收稿日期：2021-06-17;修回日期：2021-09-05;责任编辑：冯 民

基金项目：国家重点研发计划项目（2018YFB1308700）;河北省重大成果转化项目（19011826Z）;石家庄市科技计划项目（216010017A）

第一作者简介：张付祥（1973—），男，河北青县人，教授，博士，主要从事机电控制与机器人方面的研究。

E-mail：zhangfx@hebust.edu.cn

Study on character recognition algorithm for end face of bundled special steel bars

ZHANG Fuxiang1，GUO Wang1，HUANG Yongjian2，WANG Chunmei3，HUANG Fengshan1

（1.School of Mechanical Engineering，Hbebei University of Science and Technology，Shijiazhuang，Hebei 050018，China;2.Shijiazhuang Iron & Steel Company Limited，Shijiazhuang，Hebei 050031;3.School of Electrical Engineering，Hebei University of Science and Technology，Shijiazhuang，Hebei 050018，China）

Abstract：In order to realize the traceability of the whole process of special steel bar production information，the production environment and shape characteristics of special steel bar were analyzed.The marking scheme based on double mark points was adopted，and the machine vision technology was used to realize the character recognition of the end face of the bundled special steel bars.First，Hough transform was used to segment the end face image of the bundle of special steel bars into single ones.Secondly，an image enhancement algorithm based on wavelet transform was used to enhance the end face image of a single special steel bar.Then，the MSER algorithm and the edge detection algorithm were combined to complete the detection of the character area of a single special steel bar，and the character segmentation was completed based on the projection method.Finally，the end face character recognition of each special steel bar was completed by creating and training an SVM classifier，and the end face character recognition results of the bundle of special steel bars were output and saved.The results show that the new algorithm can meet the requirements of character identification in the production process of bundles，and the accuracy of character recognition can reach 97.35%.With combination of the Hough transform，the wavelet transform image enhancement algorithm，MSER algorithm，edge detection algorithm，projection method，and SVM classifier and other algorithms to character identification of special steel bar end face，the new algorithm provides reference for information acquisition，information transfer and information traceability of special steel bar production.

Keywords：

computer image processing;special steel bar;information marking;image enhancement;character segmentation;character recognition

特钢是建筑、机械、汽车等行业不可或缺的材料，其强度、硬度更大，可塑性好，中国作为钢铁产量第一大国，特钢棒材年产量在1 500万t以上。……