基于改进YOLOv7的雾天道路目标检测算法
An Improved YOLOv7-Based Algorithm for Road Target Detection in Foggy Conditions
潘宗友 1孔田增1
作者信息
- 1. 兰州石化职业技术大学 汽车工程学院,甘肃 兰州 730060
- 折叠
摘要
针对于雾天交通场景下道路目标检测的准确性亟待提升的问题,提出一种基于改进YOLOv7-tiny算法.首先,将YOLOv7-tiny中引入上采样算子CARAFE(Content-Aware ReAssembly of Features)模块,以此来提升模型的特征融合能力;将ELAN-tiny模块与多元分支模块DBB(Diverse Branch Block)融合,以增强模型提取特征信息的能力,提升目标检测精度.改进算法在保证轻量化的前提下,能够有效地提升对雾天道路目标的检测性能.实验结果表明,在模型参数量和计算量少量增加的情况下,改进算法相较于基线算法,mAP@50 和mAP@50∶95 分别提升了1.0%和0.5%,论证了改进算法的有效性.
Abstract
To address the urgent need to improve the accuracy of road target detection in foggy traffic conditions,in this paper,an improved YOLOv7-tiny algorithm is proposed.First,YOLOv7-tiny is introduced into the upsampling operator CARAFE module to improve the ability to fuse features.The ELAN-tiny module is integrated with a diverse branch block(DBB)to strengthen the model's feature extraction ability and improve detection accuracy.The improved algorithm,while maintaining a lightweight structure,can effectively enhance the detection ability of road targets in foggy conditions.Experimental results demonstrate that with a slight increase in the number of parameters and computational load,the improved algorithm achieves a 1.0%and 0.5%increase in mAP@50 and mAP@50∶95,respectively,compared to the baseline algorithm,thus validating the effectiveness of the proposed modifications.
关键词
雾天天气/道路目标检测/算法/深度学习Key words
foggy weather/road target detection/algorithm/deep learning引用本文复制引用
出版年
2024
NETL
NSTL
万方数据