SwinDet: A Multi-level RGB-D Fusion Framework for On-orbit Space Target Detection

The growing complexity of the space environment poses significant challenges to space target detection, particularly in intricate scenarios. Fine-grained perception is crucial for achieving situational awareness, yet single-modality methods are constrained by performance limitations in addressing diverse and high-precision demands. To tackle these issues, we propose a dual-stream multimodal detection algorithm utilizing the Swin Transformer as the backbone and incorporating channel and spatial fusion at each level to enhance information integration and modality complementarity. Experimental results on space-based RGB-D detection tasks demonstrate the algorithm’s robustness in complex environments and its potential for high-precision situational awareness applications.