摘要
雾天条件下,图像质量较低,导致目标检测存在难度。传统的YOLOv5s算法可以在雾天环境中进行检测,但是检测速度慢。对此,提出一种名为YOLO-FOG的雾天目标检测算法,可以提高雾天环境中的检测速度。这一算法在主干网络部分使用RepVGG结构,减小计算量,提高特征表示能力,加快推理速度,以提高检测的实时性。试验结果表明,针对RTTS数据集,自行车、公共汽车、汽车、摩托车、行人五类目标的平均精度依次可以达到81.72%、79.99%、89.24%、73.46%、83.34%,并且识别时间仅为每张0.065 s。YOLO-FOG雾天目标检测算法兼顾准确性和实时性,具有良好的应用前景。
Under foggy condition,the image quality is low,which makes object detection difficult.The traditional YOLOv5s algorithm can detect in foggy environment,but the detection speed is slow.In this regard,a foggy target detection algorithm named YOLO-FOG was proposed,which can improve the detection speed in foggy environment.This algorithm uses the RepVGG structure in the backbone network to reduce the amount of computation,improve the feature representation ability,and accelerate the inference speed,so as to improve the real-time detection performance.The experimental result shows that for the RTTS dataset,the average accuracy of the five types of targets such as bicycle,bus,car,motorcycle and pedestrian can reach 81.72%,79.99%,89.24%,73.46%and 83.34%respectively,and the recognition time is only 0.065 s per sheet.The YOLO-FOG foggy target detection algorithm has both accuracy and real-time performance,and has a good application prospect.
出处
《装备机械》
2024年第2期1-4,30,共5页
The Magazine on Equipment Machinery
关键词
雾
目标
检测
算法
Fog
Target
Detection
Algorithm
