Radar-infrared small targets fusion detection based on attention mechanism in frequency domain

This paper proposes a radar-infrared fusion method based on the frequency-domain attention mechanism (FAM) to improve the accuracy and robustness of small target (pixel ratio <1%) detection in complex environments. The features are enhanced by wavelet transform (WT) and (CFAR), and the modal weights are adaptively optimised by combining the dynamic decision-level fusion framework. On the ZJU-Multispectrum dataset, the smaller target accuracy is improved by 12.7%, the recall is enhanced by 11.8%, and the full-sample mAP50-95 is improved by 5.2%.