火山灰
遥感
地质学
火山
环境科学
矿物学
统计分析
特征(语言学)
反射率
校准
作者
Bruno Andò,Mattia Manenti,Salvatore Castorina,Luciano Sebastián Martinez-Rau,Sebastian Bader
出处
期刊:
[Institute of Electrical and Electronics Engineers]
日期:2026-05-01
卷期号:: 1-6
标识
DOI:10.1109/i2mtc66907.2026.11694713
摘要
The real-time monitoring of volcanic ash morphology is critical for improving dispersion forecasts and mitigating risks to aviation, economy and public health. This paper introduces a vision-based measurement approach for the automated characterization of volcanic ash morphology. Addressing the limitations of traditional monitoring infrastructure, the study presents a measurement system exploiting unsupervised machine learning aimed to perform shape-classifications of ash particles without the availability of labeled training data. Using a K-Means algorithm driven by three geometric features, the system autonomously partitions ash samples into statistically distinct morphological classes. The number of clusters has been selected as the one minimizing the Davies-Bouldin Index, which takes into account both the compactness of the individual clusters and their relative separation. For this index, a minimum value of 0.94 has been achieved in the case of four clusters. The morphological characterization carried out in the study confirmed that the identified clusters correspond to the following physically meaningful categories: regular-compact, regular-elongated, irregularcompact, and irregular-elongated. This work establishes the foundation for the development of autonomous, real-time sensing networks providing useful information for the refinement of dispersion forecasts and the support of air traffic management during volcanic events.
科研通智能强力驱动
Strongly Powered by AbleSci AI