基于改进YOLOv8的遮挡车辆目标检测算法研究

打开文本图片集

中图分类号：TP391.41 文献标志码：A 文章编号：2095-2945（2025）27-0034-05

Abstract：TheimprovedYOLOv8algorithmusesacombinationofPuzleMixandMosaictoreplace theoriginalalgorithm Mosaicdataenhancement method，whichimprovestheproblemoflowacuracyoftheYOLOv8algorithmindetectingoluded objects.Twoatentionmodules，CBAMandSENet，areintroducedintoBackbonetoimprovetheproblemofdificultyin accurately detecting under complex weather and mutual occusion.When calculating regression loss，the EIoU + Soft_NMSmethod isused tosolvetheproblemofmissingoccudedobjectsandslowconvergencespeed.TheimprovedYOLOv8algorithminherits the efficiency of the original algorithm and improves the detection accuracy and accuracy of vehicle targets.

Keywords： target detection; YOLOv8；attention;accuracy;blocking vehicle

车辆目标检测最重要的任务是同时解决准确率与实时性的问题，YOLO算法能够快速识别和跟踪各种物体，该算法的提出代表着实时目标检测这一技术难题的突破。（剩余4030字）