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

TL;DR

This paper presents a supervised semantic segmentation workflow using SPOT-6/7 imagery to improve burned area delineation accuracy and efficiency for emergency wildfire response, incorporating land cover data and optimization techniques.

Contribution

It introduces a novel workflow that enhances burned area delineation performance and robustness while addressing real-time constraints in emergency scenarios.

Findings

SegFormer requires more resources than U-Net for similar performance.

Land cover data improves model robustness without increasing inference time.

Test-Time Augmentation boosts delineation accuracy but increases inference time, mitigated by optimization.

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.