Improving object detection by selectively fusing radar and optical channels

Revealing Inter-Channel Modality Discriminability: Channel Selective Interaction Fusion Network for Optical-SAR Object Detection

Computer Vision and Pattern RecognitionMultimedia

Summary

Detecting objects from satellite images can be hard because radar (SAR) and optical images show different kinds of information that don’t always match well. The authors created a new method that carefully picks and combines the most useful parts from each type of image to improve detection. Their method separates common features from unique ones in each image type and reduces confusing or irrelevant data. When tested, this approach was better at finding objects than earlier methods.

What this means in practice

  • For remote sensing teams: Enhance the accuracy of object detection in satellite images by selectively combining radar and optical data channels for improved feature use.
  • For geospatial analytics companies: Develop advanced tools to detect and classify objects more reliably in multimodal satellite imagery by integrating structural and appearance cues.$Commercial implications: Enables creation of superior commercial geospatial products with better object detection from combined SAR-optical data.

Authors

Haojie Li, Yuanfei Huang, Lingfei Song, Jia Li, Lizhi Wang

Abstract

Synthetic aperture radar (SAR) and optical imagery provide complementary structural and appearance information for remote sensing object detection. However, substantial cross-modal discrepancies and indiscriminate channel interaction inevitably introduce redundant or conflicting features. Since different feature channels encode heterogeneous semantic information and contribute unequally to cross-modal fusion, selectively highlighting task-relevant channels while suppressing redundant ones is essential. To address this issue, we propose the Channel Selective Interaction Fusion Network (CSIF-Net), which performs selective channel-level cross-modal interaction. CSIF-Net first employs a frequency decomposition module to construct frequency-aware SAR inputs, facilitating the separation of structural information from noise-related responses. It then models cross-modal channel relationships through a cross-attention channel selection module and dynamically partitions the features into shared and modality-specific components. Finally, a dual-branch fusion module performs cross-modal interaction and channel realignment for shared features while independently enhancing SAR-specific features, thereby exploiting complementary information and suppressing modality interference. Extensive experiments on the OGSOD 1.0 and M4-SAR datasets demonstrate that CSIF-Net achieves state-of-the-art performance in both horizontal and oriented object detection, surpassing the existing methods by 4.01% and 3.18% gains, respectively.