基于YOLOv8的三阶段车牌检测  

Three-Stage License Plate Detection Based on YOLOv8

作　　者：卢鹏涛 卢晨雨 LU Pengtao;LU chenyu(College of Computer and Infomation Technology,China Three Gorges University Yichang,Hubei,443002;intenational College,Guangzhou College of Commerce Guangzhou,Guangdong,511363)

机构地区：[1]三峡大学计算机与信息学院,湖北宜昌443002 [2]广州商学院国际学院,广东广州511363

出　　处：《长江信息通信》2024年第4期73-76,共4页Changjiang Information & Communications

摘　　要：车辆牌照识别在智能交通运输系统中有重要意义。针对车辆牌照检测时存在假阳性的问题,文章提出一个三阶段车辆牌照检测方法来定位自然场景下的车辆牌照。所提方法主要包括三个步骤。首先,通过YOLOv8算法检测车辆位置并将其标记。其次,对标记后的车辆区域检测文本感兴趣区域以此获取待定车辆牌照样本并对其识别。最后,通过建立车辆牌照数据库并对比第二步获取的识别信息以此确定车辆牌照。通过实验结果表明,文章提出的三阶段车辆牌照检测方法能有效降低车辆牌照检测假阳性问题,在精度和速度方面达到了较好的效果,车牌检测的准确率达到97.5%,为后续的车辆牌照识别任务提供了一个有效的方法。Vehicle license plate recognition is of great significance in intelligent transportation systems.To address the issue of false positives in vchicle license plate detection,this paper proposes a threc-stage method for license plate detection in natural scenes.The proposed method consists of thrce main steps.Firstly,the YOLOv8 algorithm is used to detect the vchicle's location and mark it.Secondly,the marked vehicle regions are used to detect regions of interest for text,from which potential license plate samples are obtained and recognized.Finally,a vehicle license plate database is established,and the recognition information obtained in the second step is compared to determinc the vchiclc's liccnsc plate.Experimcntal rcsults show that the thrcc-stagc vehicle license plate detection method proposed in this paper effcctively reduces falsc positive issues in license plate detection and achieves good performance in terms of accuracy and speed,with a license plate detection accuracy of 97.5%.This provides an effective approach for subsequent vehicle license plate recognition tasks.

关 键 词：车辆牌照识别 牌照检测 YOLOv8 感兴趣区域 

分 类 号：TP393[自动化与计算机技术—计算机应用技术]