OSDMamba: Enhancing Oil Spill Detection from Remote Sensing Images Using Selective State Space Model

2025-06-22

AI Summary

To address three key challenges in remote sensing oil spill detection—severe label scarcity, extreme class imbalance between foreground (oil spills) and background, and frequent missed detections of small-scale oil patches—this paper pioneers the integration of the selective state space model (Mamba) into semantic segmentation. We propose a novel oil spill detection framework featuring an asymmetric decoder, deep supervision, a ConvSSM module, and progressive feature aggregation. This design jointly enables large-receptive-field modeling and fine-grained detail preservation, substantially improving multi-scale feature fusion and sensitivity to minority-class oil spill regions. Evaluated on two public remote sensing oil spill datasets, our method achieves new state-of-the-art performance, with mIoU improvements of 8.9% and 11.8%, respectively. These results demonstrate the effectiveness and superiority of state space models for small-object segmentation in remote sensing imagery.

Technology Category

• Computer Vision: Remote Sensing / Geospatial AI
• Machine Learning: Semi-Supervised Learning
• Search and Optimization: Learning to Search

Application Category

• Semantics and Knowledge: Methods to enhance, augment, integrate or synergize semantic models such as knowledge graphs and LLMs
• Search and Retrieval-Augmented AI: Retrieval-Augmented Generation (RAG) and multi-modal RAG
• Systems and Infrastructure for Web, Mobile and WoT: Applied ML and AI for Web-based mobile applications

Abstract

Semantic segmentation is commonly used for Oil Spill Detection (OSD) in remote sensing images. However, the limited availability of labelled oil spill samples and class imbalance present significant challenges that can reduce detection accuracy. Furthermore, most existing methods, which rely on convolutional neural networks (CNNs), struggle to detect small oil spill areas due to their limited receptive fields and inability to effectively capture global contextual information. This study explores the potential of State-Space Models (SSMs), particularly Mamba, to overcome these limitations, building on their recent success in vision applications. We propose OSDMamba, the first Mamba-based architecture specifically designed for oil spill detection. OSDMamba leverages Mamba's selective scanning mechanism to effectively expand the model's receptive field while preserving critical details. Moreover, we designed an asymmetric decoder incorporating ConvSSM and deep supervision to strengthen multi-scale feature fusion, thereby enhancing the model's sensitivity to minority class samples. Experimental results show that the proposed OSDMamba achieves state-of-the-art performance, yielding improvements of 8.9% and 11.8% in OSD across two publicly available datasets.

Problem

• Detecting oil spills in remote sensing images accurately
• Overcoming class imbalance and limited labeled samples
• Improving small oil spill area detection with global context

Innovation

• Uses Mamba's selective scanning mechanism
• Incorporates asymmetric decoder with ConvSSM
• Enhances multi-scale feature fusion