基于FFA-Net与改进YOLOv8n的雾天行车目标检测算法

Object detection algorithm for foggy driving conditions based on FFA-Net and improved YOLOv8n

针对雾天行车场景中目标检测精度差、效率低和漏检等问题,通过优化YOLOv8n网络并结合FFA-Net算法提出了一种新的雾天行车目标检测算法,以提高雾天环境下的目标检测效果。利用FFA-Net去除雾天图像中的噪声和模糊,以增强图像的清晰度和对比度。对KITTI数据集进行数据增强,并且利用迁移学习策略,以YOLOv8n模型为基础,增加小目标检测层,让网络同时关注到图像中不同尺度的特征,使网络更加关注小目标的检测,提高检测效果。在骨干网络末端引入GAM注意力机制以增强网络对全局信息的感知能力。仿真结果表明:针对雾天行车场景下的目标检测,去雾网络能将图像的清晰度做到高度还原,所提出改进算法的mAP值、Precision值和Recall值与原模型相比也分别提高4.7%、3.8%、6.4%,相比传统的目标检测算法具有更好的场景适用性和更高的检测精度。

Aiming at the problems of poor accuracy,low efficiency,and missed detections in object detection under foggy driving conditions,a new object detection algorithm was proposed by optimizing the YOLOv8n network in combination with the FFA-Net algorithm to enhance detection performance in foggy environments.The FFA-Net was used to remove noise and blur fog-affected images to enhance their clarity and contrast.Data augmentation was carried out on the KITTI dataset,and transfer learning strategies were used to integrate a small object detection layer based on the YOLOv8n model,allowing the network to simultaneously focus on features of different scales in the image.This enhancement allows for more accurate detection of small objects and boosts overall detection efficiency.A global attention mechanism(GAM)was introduced at the end of the backbone network to enhance the network′s perception of global information.The simulation experimental results show that for object detection in foggy driving scenarios,the defogging network can achieve a high degree of clarity restoration in images.The proposed improved algorithm also increases the mAP value,Precision value,and Recall value by 4.7%,3.8%,and 6.4%respectively compared to the original model.Compared with traditional object detection algorithms,this approach has better scene applicability and higher detection accuracy.