基于多级别融合的可见光和SAR图像联合识别

打开文本图片集

中图分类号：TP391.4 文献标志码：A DOI： 10.12305/j.issn.1001-506X.2026.03.09

Abstract：Toaddressthe issues ofinformationloss anddetail bluring inthefusion processofsynthetic aperture radar （SAR）andvisible lightimages，which lead to insuficient acuracy intarget detectionandrecognition，amulti-level image fusion method spanning pixel，feature，and decision levels is proposed. Firstly，perform preliminary fusion atthe pixel leveltoenhance thesensitivityof the detection model to texture and morphological details.Then，conduct further fusionat the feature level to make the model focus more on global semantic informationandcapture target features in complex backgrounds.Finaly，integrateandfusethe informationatthedecisionlevel，selecting detectionresults withhighreliability and accuracy.The model is compared and ablated experimentson a self-constructed targetrecognitiondataset，with experimental results confirming its effectiveness.

Keywords：multi-level fusion； joint recognition；synthetic aperture radar （SAR）；visible light

0引言

近年来，随着遥感技术的进步，遥感图像的分辨率不断提升，应用场景也在不断增多，目标检测和识别是其中的典型应用之一。（剩余14982字）