請用此 Handle URI 來引用此文件:
http://tdr.lib.ntu.edu.tw/jspui/handle/123456789/1320
標題: | 應用深度學習於分析颱風衛星影像 Using Deep Learning for Typhoon Satellite Imagery Analysis |
作者: | Alexander Grishin 阿里尅斯 |
指導教授: | 留長遠(Cheng-Yuan Liou) |
關鍵字: | 深度學,颱風,衛星影像, Deep Learning,Typhoon,Satellite Imagery,CNN,LSTM, |
出版年 : | 2018 |
學位: | 碩士 |
摘要: | - Tropical cyclones pose a great danger to humans, buildings, infrastructures, livestock and marine vessels, causing catastrophic damage and leading to disasters. Early warning systems with precise tools for prediction and estimation cyclone intensity can allow government to take adequate measures for informing population and managing administrative units accordingly to the threat posed. Different deterministic approaches of prediction and analysis of typhoons are applied by meteorology agencies all over the world for early detection and taking precocious measures. This work is focused on the analysis of a tropical cyclone using data-driven non-deterministic approach utilizing the latest state-of-the-art techniques of machine learning, such as Deep Learning, Convolutional Neural Nets, Long-Short Time Memory and Ensemble Learning. Three main problems are addressed in this work: tropical - extratropical cyclone classification; estimation of typhoon intensity class; estimation of central pressure of a typhoon. The results, achieved in this work showed improvements over other approaches for dealing with similar problems. Also it was shown, that the used data-driven approach can provide an objective expert system, which can detect a human-introduced biased error in labeled data and propose an objective tropical/extratropical cyclone classification decision as a useful tool and alternative to existing deterministic methods, which rely on subjective human decision. |
URI: | http://tdr.lib.ntu.edu.tw/handle/123456789/1320 |
DOI: | 10.6342/NTU201804080 |
全文授權: | 同意授權(全球公開) |
顯示於系所單位: | 資訊工程學系 |
文件中的檔案:
檔案 | 大小 | 格式 | |
---|---|---|---|
ntu-107-1.pdf | 12.73 MB | Adobe PDF | 檢視/開啟 |
系統中的文件,除了特別指名其著作權條款之外,均受到著作權保護,並且保留所有的權利。