基于深度学习的小目标林火检测实验设计与实现

打开文本图片集

中图分类号：TP391.4;S762 文献标志码：A 文章编号：1000-2006（2025）05-0235-07

Abstract：【Objective】As UAVtechnologycontinuouslydevelopsandthedeep-learning theory is explored in-depth，the applicationofUAVsinforestpatrolandforestfiremonitoringhasdrawnincreasingatention.Inthecurrentprofessional courses of electronic information such as image processing and computer vision，there isa lack of teaching cases related tosmal-targetrecognition.Also，inforestfiresampleimages，thereareproblemslikehiddenfirepointsandeasy-to-bemisseddetection.Thus，a deep-learning-basedsmal-target forest fire detectionmodel'isproposed.【Method】A lightweightbackbone network wasutilized.Moreover，aglobalatentionmechanismwas introduced inthefeature fusion layer，which help to reduce information loss and enhance the performance of the deep neural network.Additionall，a smal-targetdetectionlayeristodetectshalowerfeaturemaps，therebyachievinghigh-precisiondetectionofsmall-target forestfires.The detection effctof themodel wastestedon different datasetsand in multiple scenarios.【Result】The results indicate that the constructed model attains 84.79% in the mAP index，which is a 4.45% improvement compared withtheYOLOv5s model，andtheFPS remainsabove 6O.This verifiestherationalityand effctivenessof the model in detectingsmall-targetforestfiresfromtheaerialphotographyangle.【Conclusion】Themodel showsexcellentperformance inboth detectionaccuracyand speed.Future research can explore semi-supervised orself-supervised learning to reduce data annotation costs while maintaining recognition accuracy.

Keywords：deep learning;target detection； computer vision;smal target forest fire;model lightweighting;UAV imagery近年来，全球森林火灾频繁发生，其难以控制 性以及处置不及时等问题导致巨大的财产损失和

严重的安全隐患[1]。（剩余9944字）