基于机器视觉的铸件多区域铸字识别  

Multi-region Distributed Word-casting Recognition of the Castings Based on Machine Vision

作　　者：侯明君 孙晓龙 伍缘杰 计效园[1] 钱星云 李连杰 周建新[1] Hou Mingjun;Sun Xiaolong;Wu Yuanjie;Ji Xiaoyuan;Qian Xingyun;Li Lianjie;Zhou Jianxin(State Key Laboratory of Materials Processing and Die&Mould Technology,Huazhong University of Science.and Technology;Wuxi Branch,FAW Foundry Co.,Ltd.)

机构地区：[1]华中科技大学材料成形与模具技术国家重点实验室,武汉430074 [2]一汽铸造有限公司无锡分公司

出　　处：《特种铸造及有色合金》2022年第10期1225-1231,共7页Special Casting & Nonferrous Alloys

基　　金：国家重点研发计划资助项目(2020YFB1710100);国家自然科学基金资助项目(51905188,52275337,52090042)。

摘　　要：针对人工铸字识别方法识别效率低、工作强度大,且现有字符识别技术无法直接应用于铸字识别的问题,设计了一种基于机器视觉的铸字识别系统。该系统能对铸字图像的光强、亮度、对比度、二值化、去噪等方面进行预处理。此外,还能将预处理后的图像进行轮廓识别和铸字识别,最终获得铸字信息。针对于具有复杂结构的铸件的识别,提出了若干改进方法,在一定程度上解决了上述问题,能更好地实现对铸件全流程的生命周期管控与溯源。In view of the problems of low efficiency and high work intensity of artificial casting recognition, and failure of existing character recognition technology, a word-casting recognition system based on machine vision was designed, which made a certain breakthrough in preprocess the light intensity, brightness, contrast, binarization, denoising and other aspects of casting image. Moreover, the preprocessed image can be used for contour recognition and word-casting recognition to obtain word-casting information. Aiming at the castings with complex structure, several methods were proposed, which solved the above problems to a certain extent and could better achieve the life cycle control and traceability of the whole process of casting.

关 键 词：图像预处理 同态滤波 二值化 轮廓识别 铸字识别 

分 类 号：TP311[自动化与计算机技术—计算机软件与理论]