基于DBNet+SVTR的微电子组装电路字符识别系统  

Character Recognition System of Microelectronic Assembly Circuit Based on DBNet+SVTR

作　　者：李颖 万永 罗驰[1] 袁家军 简燕 LI Ying;WAN Yong;LUO Chi;YUAN Jiajun;JAN Yan(China Electronics Technology Group Corporation No.24 Research Institute,Chongqing 400060,China)

机构地区：[1]中国电子科技集团公司第二十四研究所,重庆400060

出　　处：《电子与封装》2025年第4期24-30,共7页Electronics & Packaging

摘　　要：微电子组装技术的发展推动组装电路的类型和产量不断增长,因此作业智能化水平亟需提升,且需对每只电路的生产状态进行实时追踪记录。基于DBNet和SVTR网络模型的电路管壳批号识别系统采用端到端的光学字符识别(OCR)模型,结合分布式技术进行相关硬件部署,可实现作业任务单智能生成、作业自动记录、作业过程智能化闭环管理等功能,提升了微电子电路生产过程的智能化水平。The development of microelectronic assembly technology has driven the continuous increase in the types and output of assembled circuits.Therefore,the level of intelligent operation urgently needs to be improved,and real-time tracking and recording of the production status of each circuit are required.The circuit tube shell batch number recognition system based on DBNet and SVTR network models adopts an end-to-end optical character recognition(OCR)model,combined with distributed technology for related hardware deployment,which can achieve intelligent generation of job task orders,automatic recording of jobs and intelligent closed-loop management of job processes,and improve the intelligence level of microelectronic circuit production processes.

关 键 词：图像处理 文本识别 文本检测 OCR DBNet SVTR 

分 类 号：TN407[电子电信—微电子学与固体电子学] TP391.4[自动化与计算机技术—计算机应用技术]