Vision Foundation Model Guided Multimodal Fusion Network for Remote Sensing Semantic Segmentation

With the rapid development of Earth observation sensors, the fusion of remote sensing (RS) data in multimodal semantic segmentation has garnered significant research focus in recent years. The fusion of multimodal data presents challenges due to discrepancies in image acquisition mechanisms among different sensors, leading to misalignment issues. To mitigate this challenge, this article presents VSGNet, a novel multimodal fusion framework designed for RS semantic segmentation. The work aims to utilize vision structure guidance derived by vision foundation model for accurate segmentation without the need for auxiliary sensors. Specifically, the framework incorporates a cross-modal collaborative network for feature embedding that blends a convolutional neural network and vision transformer to simultaneously capture both local information and long-range dependencies from the input modalities. Subsequently, a multiscale cross-modal feature fusion comprising fusion enhancement and feature recalibration modules is proposed to emphasize the adaptive multiscale interaction of diverse complementary cues between each modality while suppressing the impact of noise and uncertainties present in RS data. Extensive experiments conducted on four diverse RS datasets, i.e., ISPRS Potsdam, ISPRS Vaihingen, LoveDA, and tree mapping, demonstrate VSGNet outperforms state-of-the-art RS semantic segmentation models.