基于YOL0v8模型和Streamlit的地面小目标检测系统

打开文本图片集

中图分类号：TP391.4；TP311 文献标识码：A 文章编号：2096-4706（2025）19-0057-06

Abstract：Inground-basedaerial images，detecting smalltargets ischalengingduetotheirsmallsize，lowresolution， andsimilaritytothebackground，leadingtouboptimalpformancewithtraditionaldetectionmethods.Toanceetection eficiencyandaccuracy，andtoprovideusers withanintuitiveplatformforimageuploadandresultetrieval，thispaperdesigns a ground-based Small Target Detection system based on the YOLOv8n model and Streamlit.After comparing the YOLO performances of various YOLO versions，YOLOv8n is selected.An aerial dataset encompassng multiple ground targets is constructed and used to train the model， achievingan mAP @0.5 exceeding 90% in Small TargetDetection of theYOLOv8n model.Basedonthis，thedevelopedWebinterfaceenablestheTargetDetectionfunctionacrossimages，videosandreal-time camera nputs.Thedesigned systemoutperformstraditional methods inbothdetectionaccuracyandspeed，oferingafriendly useroperation platformwith practical application significance.

Keywords： ground-based Small Target Detection; YOLOv8; Streamlit; Target Detection; Web interface

0 引言

地面小目标检测是计算机视觉领域的关键技术，在军事侦察、环境监测、交通监管等领域具有重大战略价值。（剩余6710字）