跳转至

文章背景与核心概要

甲烷作为一种强效温室气体，是加剧全球气候变化的关键因素之一，对其进行及时有效的检测对于减缓气候变化至关重要。传统的甲烷检测流程依赖于地面密集处理，包括用于校正几何畸变的“正射纠正（Orthorectification）”以及用于增强羽流信号的“匹配滤波（Matched Filters）”，这些步骤计算繁重，极其不适合在资源受限的卫星机载端直接执行。

为了突破这一瓶颈，本文推出了 UnorthoDOS——一个专为直接利用未经正射纠正的高光谱图像训练机器学习模型而设计的新型数据集与方法。研究所提出的 U-Net 模型成功绕过了传统正射纠正和匹配滤波产品的依赖，其在全部羽流检测上的交并比（IoU）达到了 16.91%，表现接近于在正射纠正数据上训练的模型（18.47%），并大幅超越了 mag1c 匹配滤波基线模型（4.76%）。此外，作者通过采用 FP16 压缩技术验证了机载部署的可行性，该技术在将模型体积减半的同时，输出偏差控制在 0.3% 以下。

Methane Detection On Board Satellites from Unorthorectified Imagery

Summary

Timely detection of methane—a potent greenhouse gas—is crucial for effective climate change mitigation. Conventional detection pipelines rely on orthorectification (to correct geometric distortions) and matched filters (to enhance plume signals), which are designed for heavy ground processing and are poorly suited for execution onboard satellites.

This paper introduces UnorthoDOS, a novel dataset and approach designed to train machine learning models directly on unorthorectified hyperspectral imagery. The proposed U-Net models bypass both traditional orthorectification and matched-filter products, achieving performance closely matching models trained on orthorectified data (IoU of 16.91% vs. 18.47% on all plumes), while vastly outperforming the mag1c matched-filter baseline (IoU of 4.76%). Furthermore, the authors demonstrate the viability of onboard deployment by utilizing FP16 compression, which halves model size while incurring less than 0.3% output deviation.

Paper Metadata

  • arXiv Identifier: arXiv:2609.04906 [cs.CV]
  • Primary Subject: Computer Vision and Pattern Recognition (cs.CV)
  • Secondary Subjects: Artificial Intelligence (cs.AI), Machine Learning (cs.LG)
  • Submission Date: September 4, 2026
  • Authors:
  • Luca Marini
  • Maggie Chen
  • Hala Lamdouar
  • Laura Martínez-Ferrer
  • Dr. C. P. Bridges
  • Giacomo Acciarini

Abstract

As a potent greenhouse gas, methane is a major driver of climate change. Its effective mitigation relies on timely detection. Conventional detection methods rely on orthorectification to correct geometric distortions and matched filters to enhance plume signals, which are steps designed for ground processing and poorly suited to onboard execution. We introduce UnorthoDOS, a dataset and approach for training machine learning models directly on unorthorectified hyperspectral imagery, bypassing both orthorectification and matched-filter products. Our U-Net models trained on unorthorectified data approach the performance of models trained on orthorectified data (IoU 16.91% vs. 18.47% on all plumes), while both substantially outperform the mag1c matched-filter baseline (IoU 4.76%). We further demonstrate the feasibility of onboard deployment: FP16 compression halves model size with under 0.3% output deviation.

License Asset Reference

license icon