改進(jìn)YOLOv8n的塵霧環(huán)境下目標(biāo)檢測(cè)算法

打開文本圖片集

主題詞：自動(dòng)駕駛目標(biāo)檢測(cè)注意力機(jī)制多尺度特征融合塵霧環(huán)境中圖分類號(hào)：U469.6；TP242.6 文獻(xiàn)標(biāo)志碼：A DOI： 10.19620/j.cnki.1000-3703.20240036

ImprovedYOLOv8n ObjectDetection AlgorithminDust andFogEnvironment

WangZiyu’，ZhangJiancheng2，LiuYuansheng2 （1.Scholof UrbanRail TransitandLogistics，Beijing Union University，Beijing1Oo101;2.Schoolof Robotics， Beijing Union University，Beijing100101）

【Abstract】To address the issues of missed detections，false detections and lowaccuracy in detecting smalland distant objects underadverse conditions such as dustand haze，this paper proposestheEPM-YOLOv8object detection algorithm.The Eficient ChannelAtention （ECA）moduleisintegratedintotheC2f moduleof theYOLOv8nalgorithm，enablingthebackbone network to focus more effectivelyonshallowandsmallrobjectfeatures.Byadding anadditional detectionlayeranddesigning a multi-dimension feature fusion architecture，the model'sability to extracttarget featuresandits detectionaccuracyare significantlyimproved.Furthermore，alossfunctionbasedontheMinimumPointDistance IntersectionoverUnion（MPDIoU） is employedtoenhance theprecisionofboundingboxregresion.ExperimentalresultsdemonstratethattheEPM-YOLOv8model achieves a precision ratio of 83.6% and a detection accuracy of 76.8% ，exhibiting superior detection performance under challenging conditions such as haze and dust.

Key Words：Autonomous driving，Object detection，Attention mechanism，Multi-scale feature fusion，Dusty and foggy environment

【引用格式】王子鈺，張建成，劉元盛.改進(jìn)YOLOv8n的塵霧環(huán)境下目標(biāo)檢測(cè)算法[J].汽車技術(shù)，2025（6）：1-7. WANGZY，ZHANGJC，LIUYS.ImprovedYOLOv8n Object DetectionAlgorithminDustandFog Environment[J]. Automobile Technology，2025（6）： 1-7.

1前言

在塵霧環(huán)境中，圖像模糊、質(zhì)量下降導(dǎo)致有效特征提取困難，目標(biāo)檢測(cè)任務(wù)易出現(xiàn)精度降低、錯(cuò)檢和漏檢等問題。（剩余8894字）