基于YOLOv13的无人机类别检测方法

打开文本图片集

中图分类号：TP391.4 文献标识码：A文章编号：1006-8228（2025）12-14-05

Abstract：UAVdetectionfromaUAVperspectiveremainsaresearchhotspotinthefieldsoflow-atitudeeconomyandUAV autonomousobstacleavoidance.However，thereisstillalackofpubliclyavailabledatasetsspecificallydesignedforrecognizing diferentUAVmodels.Toadressthisgap，thisstudyconstructsacomprehensiveUAVimagedatasetcontainingvariousscenarios andviewpoints，coveringthreecommontypesofUAVs.Adetectionmethod integratingtheYOLOv13algorithmwithMosaicdata augmentationisproposdtoachieveeficientandaccuratemult-typeUAVidentication.Duringtraining，Mosaicdataaugmentation enhancesthemodel'sgeneralizationcapabilityComparativeexperimentsareconductedtoevaluatetheadvantagesindetection accuracyoftheproposedmethod.Experimentalresultsontheself-builtdatasetdemonstratethattheproposedmethodachievesan excellent mAP50of 95.4% ，indicating high precision whilemaintaining favorabledeployment flexibility，therebyvalidating its effectivenessin practical UAV recognition tasks.

Keywords：Object Detection;UAV;YOLOv13;Dataset;Artificial Intelligence

0引言

随着低空经济的快速发展，配备摄像头的小型无人机在日常生活记录、农业监测、安防巡查以及军事侦察等领域的应用日益广泛。（剩余7418字）