Multitask GLocal OBIA-Mamba for Sentinel-2 Landcover Mapping

2025-11-13

AI Summary

To address challenges of spatial heterogeneity, insufficient contextual modeling, and spectral ambiguity in Sentinel-2-based land use and land cover (LULC) classification, this paper proposes OBIA-Mamba—a multi-task framework integrating superpixel-driven object-based image analysis (OBIA) with global–local collaborative learning. Methodologically, superpixels serve as semantic tokens for Mamba-based sequential modeling; a dual-branch CNN–Mamba architecture is designed to jointly capture local textural details and global semantic context; and a joint loss function combining classification accuracy and boundary-aware regularization enables end-to-end optimization. Evaluated on Sentinel-2 imagery over Alberta, Canada, OBIA-Mamba achieves state-of-the-art performance, improving overall accuracy (OA) by 2.1% and Kappa coefficient by 0.023 over existing methods. Moreover, it produces classification maps with enhanced spatial coherence and finer-grained, more topologically consistent land cover patches.

Technology Category

• Machine Learning: Multi-class/Multi-label Learning & Extreme Classification
• Computer Vision: Remote Sensing / Geospatial AI
• Search and Optimization: Learning to Search

Application Category

• Systems and Infrastructure for Web, Mobile and WoT: Applied ML and AI for Web-based mobile applications
• Semantics and Knowledge: Methods to enhance, augment, integrate or synergize semantic models such as knowledge graphs and LLMs
• Search and Retrieval-Augmented AI: Multilingual and cross-lingual Web search

Abstract

Although Sentinel-2 based land use and land cover (LULC) classification is critical for various environmental monitoring applications, it is a very difficult task due to some key data challenges (e.g., spatial heterogeneity, context information, signature ambiguity). This paper presents a novel Multitask Glocal OBIA-Mamba (MSOM) for enhanced Sentinel-2 classification with the following contributions. First, an object-based image analysis (OBIA) Mamba model (OBIA-Mamba) is designed to reduce redundant computation without compromising fine-grained details by using superpixels as Mamba tokens. Second, a global-local (GLocal) dual-branch convolutional neural network (CNN)-mamba architecture is designed to jointly model local spatial detail and global contextual information. Third, a multitask optimization framework is designed to employ dual loss functions to balance local precision with global consistency. The proposed approach is tested on Sentinel-2 imagery in Alberta, Canada, in comparison with several advanced classification approaches, and the results demonstrate that the proposed approach achieves higher classification accuracy and finer details that the other state-of-the-art methods.

Problem

• Addresses Sentinel-2 landcover mapping challenges like spatial heterogeneity
• Reduces redundant computation while preserving fine-grained image details
• Balances local precision with global consistency in classification

Innovation

• OBIA-Mamba uses superpixels as tokens
• GLocal CNN-Mamba models local and global features
• Multitask optimization balances dual loss functions