基于改进YOLOv8和LPRNet的车牌识别算法  

License Plate Recognition Algorithm Based on Improved YOLOv8 and LPRNet

作　　者：宋建辉 夏彬 赵亚威 刘晓阳 SONG Jianhui;XIA Bin;ZHAO Yawei;LIU Xiaoyang(Shenyang Ligong University,Shenyang 110159,China)

机构地区：[1]沈阳理工大学自动化与电气工程学院,沈阳110159

出　　处：《沈阳理工大学学报》2025年第3期39-46,共8页Journal of Shenyang Ligong University

基　　金：辽宁省教育厅高等学校基本科研项目(LJKZ0275);沈阳市中青年科技创新人才支持计划项目(RC210247)。

摘　　要：车牌定位检测与识别是实现智能交通系统的关键技术之一。为解决传统方法定位不准确、识别率低的问题,提出一种基于改进YOLOv8和轻量级识别网络LPRNet的车牌识别算法。在YOLOv8网络架构中增加卷积块注意力模块(convolutional block attention module,CBAM),通过特征融合提升目标检测的准确率;在LPRNet中加入多头自注意力(multi head self attention,MHSA)机制,增强网络的特征提取能力;在识别网络模型的训练阶段引入对抗训练,增强模型的泛化能力和鲁棒性。在CCPD数据集上的实验结果表明,本文算法对车牌检测的平均精度均值(mAP@0.5)达到98.7%,车牌识别准确率达到97.21%,均优于其他同类主流算法。Detection and recognition of license plate location is one of the key technologies for implementing intelligent transportation systems.To solve the problems of inaccurate positioning and low recognition rate in traditional methods,a license plate recognition algorithm based on improved YOLOv8 and lightweight recognition network LPRNet is proposed.A convolutional block attention module(CBAM)is added to the YOLOv8 network architecture to improve the accuracy of object detection through feature fusion.A multi head self attention(MHSA)mechanism is added to LPRNet to enhance the network's feature extraction capability.Adversarial training is introduced during the training phase of the recognition network model to enhance its generalization ability and robustness.Experimental results on the CCPD dataset show that the mean average precision(mAP@0.5)of our algorithm for license plate detection reaches 98.7%,and the accuracy of license plate recognition reaches 97.21%,both of which are superior to other mainstream algorithms of the same type.

关 键 词：车牌识别 YOLOv8 LPRNet 对抗训练 

分 类 号：TP391.4[自动化与计算机技术—计算机应用技术]