基于EFE-LinkNet网络的SAR图像道路检测算法

打开文本图片集

中图分类号：TP753 文献标志码：A 文章编号：2095-2945（2025）28-0014-05

Abstract：Traditional SAR（SyntheticApertureRadar）imageroaddetectionalgorithms haveproblemssuchasbeinggreatly affectedbyspecklenoise，dificultyinutilizinghigh-frequencyimageinformation，andlowsegmentationaccuracy.Tosolvethe aboveproblems，thisstudyimprovestheLinknetnetworkanddesignsaroaddetectionmethodforSARimagesthatcombines efciencyadaccuracy：EFE-Linet（EficientFeatureEnhancement-Linket）Fistalcaleatureandirectioalinfoation enhancementmoduleisdesignedtoenablethenetworktocapturedetailedfeaturesanddirectional informationintheimage; secondly，afeaturefusionmoduleisdesignedtoreduceinformationlossbyeficientlyfusingdeepfeaturesanddetailedfeatures; finallyforthesampleimbalanceproblem，amixedlossfunctionisdesignedtooptimizethemodeltrainingprocessRoad detectionexperimentsarecarredoutusingrealairbomehigh-resolution SARimagedataTheresultsshowthattheproposed algorithm has good detection results and verifies the effectiveness of the algorithm.

Keywords：synthetic apertureradar；road detection；strippooling;residual element;lightweight

随着城市化进程中交通需求的不断增长，道路网络的准确检测和分析变得愈发重要。（剩余6915字）