基于OpenCV的芯片IMEI码的检测与识别  

Detection and Recognition of Chip IMEI Code Based on OpenCV

作　　者：朱德意 孙晴艺 董思凡 麻胜恒 王耀雄 高放 ZHU De-yi;SUN Qing-yi;DONG Si-fan;MA Sheng-heng;WANG Yao-xiong;GAO Fang(School of Electrical Engineering,Guangxi University,Nanning 530004,China;Guangxi CT UNITE Alpha Technology Co.,Ltd.,Nanning 530201,China;Institute of Intelligent Machines,Chinese Academy of Sciences,Hefei 230031,China)

机构地区：[1]广西大学电气工程学院,广西南宁530004 [2]广西中科阿尔法科技有限公司,广西南宁530201 [3]中国科学院合肥智能机械研究所,安徽合肥230031

出　　处：《计算机技术与发展》2022年第8期174-179,共6页Computer Technology and Development

基　　金：广西科技基地和人才专项(2020AC19253);安徽省重点研究与开发计划(202104a05020041)。

摘　　要：运动目标的检测与识别一直是计算机视觉的热点研究方向。随着计算机技术的大力发展,机器能够与运动目标检测很好地结合,有望代替人们去完成那些枯燥乏味或者是危及生命安全的工作。在芯片制造流水线中,依靠人工识别并记录芯片国际移动设备识别码(international mobile equipment identity,IMEI),效率低下且易出错。为实现芯片IMEI码的自动识别,该文使用OpenCV对视频进行预处理,通过灰度化提取芯片上的全部轮廓,通过膨胀、腐蚀操作处理轮廓,再通过面积和周长两个参量筛选得到芯片的IMEI码区域,进一步通过将该区域与原视频结合得到含有芯片IMEI码的视频,最后使用Tesseract-OCR识别芯片IMEI码。同时做了基于神经网络CRNN进行识别的对比实验,首先通过生成器函数构造了一个由字母、数字和冒号组成的数据集用于模拟芯片IMEI码输入CRNN网络进行预训练,在此基础上用15张IMEI码的图片进行迁移学习,最后基于学习到的新模型对视频中的三块芯片进行识别。通过对比发现,在小样本数据集的情况下,该方法识别芯片IMEI的准确率远远超过CRNN识别芯片IMEI码的准确率。The detection and recognition of moving targets has always been a hot research topic of computer vision.With the vigorous development of computer technology,machines can be well integrated with moving target detection to hopefully replace people to complete those boring or life-threatening tasks.In the chip manufacturing pipeline,the identification and recording of the chip international mobile equipment identity(IMEI)is finished manually,which is inefficient and error-prone.In order to realize the automatic identification of the chip IMEI code,we use OpenCV to preprocess the video,extract all the contours on the chip through grayscale lithography,process the contours through expansion and corrosion operations,and then filter through the two parameters of area and perimeter to obtain the chip’s IMEI code region,which is further combined with the original video to obtain a video containing the chip IMEI code.Finally,Tesseract-OCR is used to identify the chip IMEI code.At the same time,a comparison experiment of identifying IMEI codes with the CRNN network is performed.A data set composed of letters,numbers and colons is constructed through the generator function and fed into the network for pre-training.Transfer learning is used to transfer the knowledge learning from the above artificial dataset to a small dataset containing 15 real IMEI code images.The optimized model is used to identify the IMEI codes of the three chips in the video.It is found that the accuracy rate of the proposed OpenCV-based method to identify IMEI codes exceeds that of the CRNN network in case of small training datasets.

关 键 词：IMEI OPENCV Tesseract-OCR 运动目标 字符识别 CRNN 

分 类 号：TP510[自动化与计算机技术] TP751