针对暗光环境下无人机跟踪的图像增强方法

打开文本图片集

中图分类号：TN911.73-34；TP389.1 文献标识码：A 文章编号：1004-373X（2025）15-0057-07

Imageenhancement method forUAV trackingin dark light environment

CHENAnlan1²，XIAO Yufeng¹，HEFeiyu³，HE Tao1，² （1.School of Mechanical Engineering，Hubei Universityof Technology，Wuhan 43Oo68，China;

2.HubeiKeyLaboratoryofModernManufacturingQualityEngineering，Wuhan43Oo68，China;

3.Hubei Institute of Quality Supervision and Inspection，Wuhan 43Oooo，China）

Abstract：UAVtracking tasksareoftenperformedinadarklightenvironment.Theexisting trackersarerarelytrained to adapttotheneedsofthedarklightenvironment，whichresultsinasignificantdeclineinthesuccessateandaccuracyrateof thetrackerinthedarklightenvironent.Inthispaper，ageneraldarklightimageenhancementmethodisproposedwhichcan enhancethecapturediagebeforetheimageenteringthetracker，sothatthetrackercanmaintaingoodtracking performance n thedarklightenvironment.OnthebasisoftheRetinextheory，animageprocesingmethodisdesigned.Inthispaper，a lightweightimage procesing networkcalld ME-Netisadopted.TheME-Netcanprocesslightmapsandnoise maps iteratively toimproveimagequality，andbyimprovingtheSSRfunction，thedatalosscausedbygain-ofsetisreducedeectively.Inder tovalidateteectivessoftdrk-lightiageitesiferroposedxprimentsreouctedovraloplarak ontheUAVDark135dataset.Theexperimentalresultsshowthattheimproveddark-lightintensifiersignificantlyimproves the successrateandacuracyrateofthetrackerunderdarklightconditions，andalsohasdiferentdegresofimprovementin comparisonwithtoseoftheothercommonlyuseddark-light intensifiers.Tosumup，theefectivenessoftheimproveddark-light intensifierhasbeenverified.

Keywords：Retinex；single-scaledRetinex;darklightenvironment;imageenhancement；objecttracking;lightweight network;UAV;Siamesenetwork

0 引言

视觉目标跟踪是当前计算机视觉研究的一个热点，已使用于许多与无人机（UAV）相关的应用中。（剩余10242字）