Answers are grounded in this paper and cite their sources.
Optical–synthetic aperture radar (SAR) object detection can obtain robust and high-precision performance by integrating the rich semantic information in optical imagery with the stable structural information in SAR imagery. However, harsh weather conditions and electromagnetic interference environments often lead to a rapid degradation in imaging quality or even data loss, which inevitably imposes a formidable challenge on achieving robust cross-modal feature fusion. To alleviate this issue, an adaptive spatial–frequency interaction fusion framework with competitive distillation (ASIF-CD) is proposed in this article. First, to mitigate the distribution differences between optical and SAR modalities, the adaptive spatial–frequency interaction module is developed to achieve cross-modal feature alignment and simultaneously suppress background noise. Then, the dynamic contribution fusion module, which balances the contributions of optical spectral information and SAR structural features, is adopted to adjust feature fusion weights. Subsequently, a multimodal competitive calibration distillation strategy is designed, which constructs a dynamic competitive optimization loss function to guide the fusion branch in achieving knowledge learning and mitigating the degradation of fused features. Finally, extensive experiments on the OGSOD and M4-SAR public datasets demonstrate that ASIF-CD yields mAP₅₀ scores of 92.6% and 86.0%, respectively, which consistently outperforms existing state-of-the-art multimodal fusion object detection methods.
No takes yet. Share an insight, caveat, or question.
Liu et al. (2026) studied this question.
Synapse has enriched 10 related papers. Consider them for comparative context: