低能见度下GYOLOv5-SPD算法的车牌检测  

License plate detection in low visibility based on GYOLOv5-SPD algorithm

作　　者：李泽 李小龙[1,2,4] 杨忠祥 谭永滨 LI Ze;LI Xiaolong;YANG Zhongxiang;TAN Yongbin(School of Surveying and Geoinformation Engineering,East China University of Technology,Nanchang,Jiangxi 330013,China;CNNC Engineering Research Center of 3D Geographic Information,East China University of Technology,Nanchang,Jiangxi 330013,China;College of Resource Engineering,Yunnan Tourism College,Kunming,Yunnan 650221,China;Key Laboratory of Mine Environmental Monitoring and Improving around Poyang Lake,Ministry of Natural Resources,Nanchang,Jiangxi 330013,China)

机构地区：[1]东华理工大学测绘与空间信息工程学院,江西南昌330013 [2]东华理工大学中核三维地理信息工程研究中心,江西南昌330013 [3]云南旅游学院资源工程学院,云南昆明650221 [4]自然资源部环鄱阳湖矿山环境监测与治理重点实验室,江西南昌330013

出　　处：《北京测绘》2024年第5期655-660,共6页Beijing Surveying and Mapping

基　　金：国家自然科学基金(4226107842361067);江西省重点研发计划(20223BBE51030);江西省地质局科技研究项目(2022JXDZKJKY08)。

摘　　要：车牌自动检测技术是构建智慧城市、加强交通管理等方面的重要内容。目前车牌检测技术正在逐步完善,但对于光照条件过低、雨天、雪天等能见度极低的气候环境下所拍摄的车牌照片,车牌检测技术还处于一个相对落后的水平。本文在中国城市停车数据集(CCPD)中能见度极低的车牌图像的基础上利用添加了伽玛(Gamma)变换的你只需看一次(YOLO)v5s模型与空间到深度层及无步长卷积层(SPD-Conv)进行融合用于车牌检测,将低能见度条件下的车牌视为小物体进行检测,目的是最大限度地识别其特征以提高精度。实验结果表明,本文用到的方法在车牌定位检测阶段达到了99.68%的召回率,在低能见度条件下对车牌的定位检测相较其他算法确实有一定的优势。Automatic license plate detection technology is an important element in building smart cities and strengthening traffic management.At present,license plate detection technology is gradually improving,but it is still at a relatively backward level facing license plate photos taken in extremely low visibility conditions such as low light conditions and rainy or snowy weather.This paper fused the YOLOv5s model with Gamma transform with space-to-depth(SPD)-layer followed by a non-strided convolution(Conv)for license plate detection based on license plate photos taken in extremely low visibility in the Chinese City Parking Dataset(CCPD)and treated license plates in low visibility conditions as small objects for detection,so as to identify their features to maximize accuracy.The experimental results show that the method used in this paper achieves a recall rate of 99.68%in the license plate location detection phase,and the location detection of license plates in low visibility conditions does have some advantages over other algorithms.

关 键 词：车牌检测 SPD-Conv Gamma变换 GYOLOv5-SPD 低能见度 

分 类 号：P258[天文地球—测绘科学与技术]