摘要
为解决复杂环境下车牌检测与识别模型速度慢、精度低的问题,提出了一种复杂环境下能高精度进行车牌检测和识别的端到端车牌识别算法。在YOLOv7网络层的特征层输出的过程中加入了CBAM通道注意力机制,提高了模型的特征提取能力;采用改进后的YOLOv7算法对复杂环境下的车牌进行检测,将检测到的车牌区域进行预处理操作,将经过处理的车牌输入到改进的CNN识别模型进行字符识别。实验结果表明,加入注意力机制后的YOLOv7检测模型的检测均值平均精度达到87.5%,改进后的模型识别准确率达到97.16%,明显优于传统的车牌识别技术,且在复杂环境识别效果良好,具有实际应用价值。
In order to solve the problems of slow license plate detection and recognition models and low detection accuracy in complex environments,an end⁃to⁃end license plate recognition algorithm that can detect and recognize license plates with high ac⁃curacy in complex environments is proposed.In the feature layer output process of the YOLOv7 network layer,the CBAM channel attention mechanism is added to improve the feature extraction capability of the model;the improved YOLOv7 algorithm is used to detect license plates in complex environments,and the detected license plate areas are pre⁃processed.In the processing operation,the processed license plate is input into the improved CNN recognition model for character recognition.Experimental results show that the average detection accuracy of the YOLOv7 detection model after adding the attention mechanism reaches 87.5%,and the recognition accuracy of the improved recognition model reaches 97.16%,which is significantly better than the traditional license plate recognition technology,and the recognition effect is good in complex environments,has practical application value.
作者
吴媛媛
石琦
Wu Yuanyuan;Shi Qi(School of Computer Science and Information Engineering,Hubei University,Wuhan 430000,China;School of Artificial Intelligence,Hubei University,Wuhan 430000,China)
出处
《现代计算机》
2024年第17期7-12,共6页
Modern Computer
关键词
深度学习
目标检测
车牌识别
注意力机制
YOLOv7
deep learning
target detection
license plate recognition
attention mechanism
YOLOv7
作者简介
通信作者:吴媛媛(1999-),女,河南信阳人,硕士研究生,研究方向为目标检测,E⁃mail:1641940450@qq.com;石琦(2000-),女,湖北黄冈人,硕士研究生,研究方向为图像处理。
