Enhancing deep learning performance on burned area delineation from SPOT-6/7 imagery for emergency management

2025-12-12

AI Summary

Accurate and rapid delineation of burned areas (BA) post-wildfire is critical for emergency response and ecological restoration, yet existing remote sensing methods often neglect real-time processing and robustness requirements. This paper addresses emergency-scenario constraints by proposing a lightweight semantic segmentation framework tailored for high-resolution SPOT-6/7 imagery. Methodologically, it introduces land cover classification as an auxiliary task within a multi-task learning paradigm—enhancing model generalization under limited training samples without incurring additional inference latency. Furthermore, it systematically evaluates the synergistic optimization of test-time augmentation (TTA) and mixed-precision inference, effectively offsetting TTA’s computational overhead. Experiments demonstrate that SegFormer and U-Net achieve comparable Dice/IoU scores under small-sample conditions; multi-task learning notably improves robustness against domain shifts and label scarcity. The integrated pipeline satisfies both stringent real-time constraints and high spatial accuracy requirements for operational BA mapping.

Technology Category

• Computer Vision: Segmentation
• Machine Learning: Multi-class/Multi-label Learning & Extreme Classification
• Intelligent Robots: Multimodal Perception & Sensor Fusion

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

After a wildfire, delineating burned areas (BAs) is crucial for quantifying damages and supporting ecosystem recovery. Current BA mapping approaches rely on computer vision models trained on post-event remote sensing imagery, but often overlook their applicability to time-constrained emergency management scenarios. This study introduces a supervised semantic segmentation workflow aimed at boosting both the performance and efficiency of BA delineation. It targets SPOT-6/7 imagery due to its very high resolution and on-demand availability. Experiments are evaluated based on Dice score, Intersection over Union, and inference time. The results show that U-Net and SegFormer models perform similarly with limited training data. However, SegFormer requires more resources, challenging its practical use in emergencies. Incorporating land cover data as an auxiliary task enhances model robustness without increasing inference time. Lastly, Test-Time Augmentation improves BA delineation performance but raises inference time, which can be mitigated with optimization methods like Mixed Precision.

Problem

• Enhances deep learning for burned area delineation from SPOT-6/7 imagery
• Addresses efficiency and applicability in emergency management scenarios
• Evaluates model performance, resource use, and optimization for practical deployment

Innovation

• U-Net and SegFormer models for semantic segmentation
• Land cover data as auxiliary task for robustness
• Test-Time Augmentation with Mixed Precision optimization