摘要
随着深度学习的不断发展,汽车自动驾驶已成为一种趋势,自动驾驶的安全问题是最重要的。其中,能准确识别复杂环境下密集的交通指示牌是保障安全驾驶的一个重要环节,针对目前检测模型对交通指示牌召回率不够高的问题,在YOLOv5的基础上提出了YOLOv5-ACB。经过300次的迭代训练,实验结果表明YOLOv5-ACB模型的mAP为62.9%、mAP50为83.6%、召回率为76.6%,相比原始的YOLOv5模型的mAP为62.45%、mAP50为82.6%、召回率为74.6%,均有较好的提升,说明所提出的改进模型降低了交通指示牌的错检和漏检率。
With the continuous development of deep learning,autonomous driving of cars has become a trend.And the safety of autonomous driving is undoubtedly the most important issue,among which the ability to identify dense traffic signs accurately in complex environment is critical to ensure safe driving.After 300 iterations of training,experimental results show that the proposed YOLOv5-ACB model has an mAP of 62.9%,an mAP50 of 83.6%and a recall rate of 76.6%,which are better than the original YOLOv5 model with an mAP of 62.45%,an mAP50 of 82.6%and a recall rate of 74.6%.This indicates that the proposed improved model reduces the error and miss detection rates of traffic signs.
作者
黄豪杰
唐宗璐
杨敏
李航
HUANG Haojie;TANG Zonglu;YANG Min;LI Hang(School of Computer and Information Engineering,Nanning Normal University,Nanning 530100,China;Beihai Foreign Language Experimental School,Beihai 536000,China;Hepu County Experimental School,Beihai 536100,China;School of Artificial Intelligence,Guangxi University for Nationalities,Nanning 530006,China)
出处
《无线电通信技术》
2023年第4期616-621,共6页
Radio Communications Technology
