基于深度学习的锂电池表面字符识别和缺陷检测
Surface Character Recognition and Character Defect Detection of Lithium Bat-teries Based on Deep Learning
刘明尧 1索广飞1
作者信息
- 1. 武汉理工大学 机电工程学院,武汉 430070
- 折叠
摘要
该文针对在锂电池的生产过程中,软包锂电池表面喷码字符识别和缺陷检测,由于人工检测耗时长、成本高等缺点,提出了基于CnOCR的字符识别方法和基于改进YOLOv8模型的字符缺陷检测方法.该方法首先利用CnStd算法对字符区域进行了定位,利用YOLOv8模型对字符进行训练,检测出有缺陷的字符.根据字符区域特点进行图像增强、二值化和字符分割等处理,采用CnOCR模型进行字符的识别.深度学习方法提高了字符识别和缺陷检测的准确率,并且保证了整个检测系统的识别和检测速度.实验结果表明,字符识别率在96%以上,字符缺陷检测率在94%以上,符合锂电池自动化生产线的生产需要.
Abstract
Aiming at the shortcomings of manual detection in character recognition and defect detection of soft-pack lithium battery surface coding,a character recognition method based on CnOCR and a character defect detection method based on improved YOLOv8 model are proposed.The method first uses the CnStd algorithm to locate the character area,and uses the YOLOv8 model to train the characters to detect defective characters.According to the characteristics of the character area,image enhancement,binarization and character segmentation are carried out,and the CnOCR model is used for character recognition.The deep learning method improves the accuracy of character recognition and defect detection,and ensures the recognition and detection speed of the entire detection system.The experimental results show that the character recognition rate is above 96%and the character defect detection rate is above 94%,which meets the production needs of the lithium battery automated production line.
关键词
软包锂电池/字符识别/字符缺陷检测/CnOCR/YOLOv8神经网络/电池自动化设备Key words
soft pack lithium battery/character recognition/character defect detection/CnOCR/YOLOv8 neural net-work/battery automation equipment引用本文复制引用
出版年
2024
NETL
NSTL
万方数据