基于多模态特征低秩融合的遥感地物要素分类

打开文本图片集

中图分类号：TP391 文献标志码：A DOI：10.3969/j.issn.1673-3819.2025.04.010

引用格式：，，，等.基于多模态特征低秩融合的遥感地物要素分类[J].指挥控制与仿真，2025，47（4）：65-73.LIUWJ，WUXDONGFetalLand-coverclasificatiowithemotesnsingimagesbasedonlowrankfusionofultidalea-tures[J].Command Control& Simulation，2025，47（4）：65-73.

Land-cover classification with remote sensing images based on low-rank fusion of multimodal features

LIU Wenjie，WU Xiaoning†，DONG Fuan，ZHANG Jinwen，LI Yiyang，CHEN Yong （NorthAutomatic Control TechnologyResearch Institute，Taiyuan O3OoO6，China）

Abstract：Multimodalremote-sensinglandclassificationaims toachievemoreaccurateandcomprehensiveextractionof land features inremotesensingimagesbyintegratingfeatureinformationfrommultipleremotesensingdatasources.Thisarticle proposesaunified multimodalremote sensing feature clasification network，which includes：a weight sharing backbone network responsibleforextracting preliminary featurerepresentationsfromtheinputdataofeach modality；Themultimodal featurelowrank fusionmoduleperformscrossmodal transmissiononhigh-level semanticfeatures toenhancesemantic interactionbetweenmodalities；Theupsamplingoperationisresponsibleforrestoring thefusedfeaturemaptothesameresolution as the input image. This algorithm achieved 91. 23% OA and 83.28% mIoU in remote sensing land feature classificationtasks，fectivelyallviatingtheproblemsofinsuffiientaccuracyndinsuficientutilizaionofmultimodalinfotion facedbytraditionalsinglemodal remotesensing clasificationmethods throughfeaturelowrank fusion technology，thereby significantlyimproving theperformanceofland featureclassification.

KeyWords：remote-sensing images；land-cover classification；multimodal learning；data fusion

遥感地物要素分类旨在对遥感影像进行精细化分类处理，以区分并标识出图像中每一个像素或独立对象所属的具体类别。（剩余15082字）