Search papers, labs, and topics across Lattice.
The paper introduces BRSMamba, a novel boundary-aware network for vegetation segmentation in remote sensing imagery that addresses limitations of CNNs and Transformers in capturing long-range dependencies and boundary information. BRSMamba integrates a boundary-subject fusion perception module to extract boundary features and a boundary-body resolution module to inject boundary awareness into a noncausal state-space duality (NC-SSD) state-transition matrix. Experiments on four benchmark datasets demonstrate that BRSMamba achieves state-of-the-art segmentation accuracy with high efficiency, balancing performance and computational cost.
Achieve state-of-the-art segmentation accuracy on satellite imagery with a Mamba-based architecture that uses 10x fewer parameters.
Vegetation, as a critical ecological feature and irreplaceable material resource of the Earth's surface, plays a crucial role in environmental protection, resource assessment, and urban planning. The rapid advancement of remote sensing platforms and sensor technologies has facilitated the acquisition of high-resolution remote sensing imagery, providing excellent conditions for the detailed analysis of vegetation. However, vegetation segmentation in remote sensing imagery poses distinct challenges, including extreme scale variance, spectral ambiguity, and complex boundary characteristics. The performance of convolutional neural network-based methods in vegetation segmentation is constrained by their limited ability to model long-range spatial dependencies. In contrast, although Transformer-based architectures are effective at capturing global context, their quadratic complexity makes them impractical for processing high-resolution remote sensing images. Recently, state-space models (SSMs) have emerged as a promising alternative due to their linear complexity and efficient long-range modeling capabilities. Nevertheless, existing vision SSMs have two critical limitations: 1) insufficient modeling of boundary information and 2) the limited performance improvements offered by multidirectional scanning strategies while increasing computational cost. To address these limitations, we propose BRSMamba, a boundary-aware network that integrates two novel modules with noncausal state-space duality (NC-SSD) to enhance both boundary preservation and global context modeling. Specifically, the boundary-subject fusion perception module extracts robust boundary features via a Laplacian-of-Gaussian convolutional block and fuses them with category-centric semantics to generate a boundary-subject map. This map then directs the boundary-body resolution module to inject boundary awareness into the NC-SSD state-transition matrix, enabling selective scanning that preserves fine details and captures long-range dependencies. Experiments on four benchmark datasets demonstrated that BRSMamba achieved 78.86% mean intersection over union on the Vaihingen dataset, 82.84% on the Potsdam dataset, 54.37% on the LoveDA dataset, and 72.24% on the GID dataset while using only 10.61M parameters and 26.28G floating-point operations. These results validate the effectiveness of the proposed method in balancing the accuracy and efficiency of complex vegetation segmentation tasks.