摘要
大雾天气会导致视觉传感器成像模糊,环境感知准确率低,因而严重威胁无人驾驶的安全性能。因此,本文提出了将FFA-Net去雾算法与YOLOv5目标检测算法相匹配,进而实现雾天环境下的行车障碍检测方案。实验表明,与直接利用YOLOv5对雾天交通图像进行检测的方法相比,利用本文提出的检测方法得到的Precision值、mAP值和Recall值均有明显提高,且该方法通过图像去雾后能够更好地识别交通场景中的障碍物信息,对于提高不良天气下的行车障碍检测性能,保障无人驾驶安全具有重要应用价值。
Foggy weather can lead to blurred images of visual sensors and low accuracy of environmental perception,which severely threatens the safety of unmanned vehicles.Therefore,a scheme that matching the FFA-Net defogging algorithm with the YOLOv5 target detection algorithm to realize the detection of driving obstacles in foggy weather is proposed.Experiments show that,the precision,mAP and Recall values obtained by the proposed detection method are significantly enhanced,compared with the method using YOLOv5 solely.Furthermore,the method can better identify traffic obstacles after image defogging,which has significant application value for the detection of driving and the safety of unmanned driving.
作者
赵世吉
张金钊
林立飞
燕伟杰
Zhao Shiji;Zhang Jinzhao;Lin Lifei;Yan Weijie(School of Transportation,Shandong University of Science and Technology,Qingdao 266590)
出处
《中阿科技论坛（中英文）》
2022年第9期141-144,共4页
China-Arab States Science and Technology Forum
基金
山东科技大学研究生科研创新项目“雾天环境下视觉传感器行车障碍检测性能提升方法研究”。
