基于改进SLIC的光伏热斑超像素分割算法研究

打开文本图片集

中图分类号：TP391.4；TN219 文献标识码：A 文章编号：2096-4706（2025）23-0101-05

Abstract：Itisdiffculttoachevefastandacuratesegmentationofirregulartargets ininfrared images withlowcontrast andcomplexbackgroud，whichisacommonprobleminthefieldofinformationprocesing.Basedontheaplicationbackground ofintelligentoperationand maintenanceofphotovoltaic power station，this paper proposes asuperpixel segmentationalgorithm basedonimproved SLIC，focusingonoptimizing thecomprehensiveperformanceofthealgorithm indata procesing efficiency andboundaryrecognitionacuracy.Throughtheadaptiveseedinitialzationguidedbythegradient，thehighgradientedgeoffset iseectivelyavoded，ndtelocaldyamicompactfctorissedtoadaptelyadjusttespatialandgrayscaleweighttoftthe hot spotcontour.Basedonmult-siteeal-timeinfrareddata，forirrguarotspots，teDicecoeffcientoftheproposdmethod is0.9693，and theUnder-segmentationError（UE）isreduced to 0.0386 ，achievingboundaryfidelityandstability.Thealgorithm maintainsthesametimecomplexityas thestandardSLIC.Althoughtheaverageexecutiontimeincreasesslightly，theimprovement ofsegmentation accuracy is more significant， making it suitable for the practicalapplicationofpower station inspection.

Keywords： infrared thermal imaging;faultdiagnosis;gradient guidance;adaptive initialization;inteligentoperationand maintenance

0 引言

在全球能源结构向清洁低碳化转型的背景下，光伏发电已成为实现“碳达峰、碳中和”目标的核心驱动力[1。（剩余7384字）