To address the unmanned aerial vehicle localization problem in global navigation satellite system(GNSS)-denied environments

a visual positioning framework based on multi-source image feature learning is proposed.The method employs a lightweight Siamese network with a three-dimensional attention mechanism to resolve the heterogeneity between unmanned aerial vehicle and satellite images

enhances positional mapping through a cell division strategy

and improves matching efficiency with a confidence evaluation mechanism.Experimental results on real-world datasets demonstrate that the proposed method significantly improves localization accuracy and real-time performance compared to conventional approaches.The conclusion shows that the framework effectively overcomes the limitations of GNSS-based localization in complex environments

providing a reliable vision-based solution for unmanned aerial vehicle positioning in GNSS-denied areas.