![]() |
|
以作者查詢圖書館館藏 、以作者查詢臺灣博碩士 、以作者查詢全國書目 、勘誤回報 、線上人數:114 、訪客IP:18.218.231.89
姓名 黃世宇(Shi-Yu Huang) 查詢紙本館藏 畢業系所 資訊工程學系 論文名稱 基於VIT及向日葵8號氣象衛星台灣區域雨量預測之可行性評估
(Feasibility Assessment of Taiwan Regional Rainfall Prediction: Using VIT-based model and Himawari-8 Meteorological Satellite)相關論文 檔案 [Endnote RIS 格式]
[Bibtex 格式]
[相關文章]
[文章引用]
[完整記錄]
[館藏目錄]
至系統瀏覽論文 (2026-8-1以後開放)
摘要(中) 本篇論文使用基於transformer的深度學習模型TSViT(Temporal Spatial Vision Transformer[3])以及其他深度學習模型,以三種面向探討在只使用向日葵8號的衛星光譜影像的情況下,降雨量預測之可行性。資料為2017年6月其中十天之向日葵8號衛星影像以及交通部中央氣象局劇烈天氣監測系統QPESUMS產品之歷史一小時降雨資料。
由於現今的降雨預測產品大多需要許多的資料做為產品內預測模型的輸入,包含但不限於各類氣象衛星影像、地表雷達回波資料、各地區氣象站所蒐集的資料以及歷史資料等等,並且需要多位專業氣象人員依據不同的天氣狀況選擇合適的降雨預測產品和模型輸入,整個流程耗時費力。如能單只使用氣象衛星光譜影像就可獲得相對準確的降雨量,將會節省許多人物力於此。
本文依據手上的資料做了三種實驗(1)一般化模型(2)近即時模型-相同空間相近時間以及(3)近即時模型- 相同時間相鄰空間。就實驗結果而言,僅(2)具較高之可行性。摘要(英) This study investigates the feasibility of rainfall prediction using only Himawari-8 satellite spectral imagery. The research explores the potential of the deep learning model TSViT (Temporal Spatial Vision Transformer[3]) along with other deep learning models in three different aspects. The utilized dataset consists of Himawari-8 satellite images and corresponding ground-based radar-detected rainfall distribution maps from ten days in June 2017, obtained from the Central Weather Bureau of Taiwan.
Current rainfall prediction products typically require multiple data sources as inputs for the prediction models, including various meteorological satellite images, ground-based radar data, data collected from meteorological stations, and historical data. These products also rely on the expertise of meteorological professionals to select suitable rainfall prediction products and model inputs based on different weather conditions, resulting in a time-consuming and labor-intensive process. If satisfactory rainfall predictions can be achieved using only meteorological satellite spectral imagery, it would significantly save human resources.
In this study, three experiments were conducted: (1) a generalized model, (2) a near real-time model with spatial proximity and similar time, and (3) a near real-time model with temporal proximity and adjacent space. The experimental results indicate that only the near real-time model with spatial proximity and similar time (Experiment 2) shows higher feasibility in rainfall prediction.關鍵字(中) ★ 降雨預估 關鍵字(英) ★ rainfall prediction 論文目次 第一章 緒論........................................................................................................................... 1
1-1 實驗動機及相關研究................................................................................................ 1
1-2 實驗目的: ................................................................................................................. 3
第二章 資料........................................................................................................................... 4
2-1 向日葵8號氣象衛星多光譜影像.............................................................................. 4
圖1 : 向日葵8號衛星光譜影像之各波段細節[4]...................................................... 7
2-2 劇烈天氣監測系統QPESUMS產品之降雨量資料..................................................... 7
2-3 資料時間、空間解析度、資料日期以及細節: ............................................................ 8
圖2 : 降雨量分布圖。............................................................................................ 11
第三章 模型簡介 Temporal Spatial Vision Transformer (TSViT)......................................... 11
3-1 TSViT..................................................................................................................... 11
3-2 選擇如TSViT基於Transformer模型架構之原因: ............................................... 12
3-3 TSViT模型架構簡介............................................................................................... 12
圖3 : TSViT架構簡圖與ViT的Transformer Encoder架構簡圖。........................ 13
圖4 : TSViT的多個cls tokens。........................................................................... 14
圖5 : TSViT的詳細架構。圖片取自[3]。............................................................... 16
3-4 TSViT 針對時序光譜影像分類/語意分割任務的特殊設計&inductive bias .............. 17
3-5 光譜影像與TSViT 模型訓練收斂性的討論.............................................................. 19
第四章 實驗設計、結果與討論............................................................................................. 20
4-1 一般化模型: ........................................................................................................... 20
圖6 : U-Net一般化模型實驗結果之mIoU前三好雨量預測圖。........................... 23
圖7 : U-Net一般化模型實驗結果之mIoU前三遭雨量預測圖。........................... 24
4-2 近即時模型 - 相同空間相近時間............................................................................. 25
圖8 : TSViT近即時模型 - 相同空間相近時間實驗結果之mIoU前三好雨量預測圖。............................................................................................................................. 30
圖9 : TSViT近即時模型 - 相同空間相近時間實驗結果之mIoU前三遭雨量預測圖。............................................................................................................................. 31
圖10 : ConvLSTM近即時模型 - 相同空間相近時間實驗結果之mIoU前三好雨量預測圖。................................................................................................................... 33
圖11 : ConvLSTM近即時模型 - 相同空間相近時間實驗結果之mIoU前三遭雨量預測圖。................................................................................................................... 34
4-3 近即時模型 - 相同時間相鄰空間.................................................................................... 36
圖12 : 近即時模型 - 相同時間相鄰空間訓練、測試資料劃分................................. 37
圖13 : modified_ViT近即時模型 - 相同時間相鄰空間實驗結果之mIoU前三好雨量預測圖。............................................................................................................... 41
圖14 : modified_ViT近即時模型 - 相同時間相鄰空間實驗結果之mIoU前三遭雨量預測圖。............................................................................................................... 42
圖15 : CNN近即時模型 - 相同時間相鄰空間實驗結果之mIoU前三好雨量預測圖。............................................................................................................................. 44
圖16 : CNN近即時模型 - 相同時間相鄰空間實驗結果之mIoU前三遭雨量預測圖。............................................................................................................................. 45
第五章 結論......................................................................................................................... 46
參考資料與文獻................................................................................................................... 48參考文獻 [1]交通部中央氣象局 雷達迴波 https://www.cwb.gov.tw/V8/C/W/OBS_Radar.html
[2]蔡政達, et al. 使用深度學習方法與向日葵8號多頻道資料進行台灣區域衛星降雨推估 2020 Conference on Weather Analysis and Forecasting, Taipei City, Taiwan.
[3] Tarasiou, Michail, Erik Chavez及Stefanos Zafeiriou. 「ViTs for SITS: Vision Transformers for Satellite Image Time Series」. arXiv, 2023 04/14.
[4]圖片取自:https://zh.wikipedia.org/zh-tw/%E5%90%91%E6%97%A5%E8%91%B58%E8%99%9F
[5] Dosovitskiy, Alexey, Lucas Beyer, Alexander Kolesnikov, Dirk Weissenborn, Xiaohua Zhai, Thomas Unterthiner, Mostafa Dehghani, 等. 「An Image Is Worth 16x16 Words: Transformers for Image Recognition at Scale」. arXiv, 2021 06/03 [6]M. Min, C. Bai, F. Sun, C. Liu, F. Wang, H. Xu, S. Tang, B. Li, D. Di, L. Dong, and J. Li, “Estimating Summertime Precipitation from Himawari-8 and Global Forecast System Based on Machine Learning,” IEEE Trans. Geosci. Remote Sens., vol. 57, no. 5, pp. 2557-2570, May. 2019.
[7]M. Guarascio, G. Folino, F. Chiaravalloti, S. Gabriele, A. Procopio, and P. Sabatino, “A Machine Learning Approach for Rainfall Estimation Integrating Heterogeneous Data Sources,” IEEE Trans. Geosci. Remote Sens., DOI: 10.1109/TGRS.2020.3037776.
[8]Q. Zhao, Y. Liu, W. Yao, and Y. Yao, “Hourly Rainfall Forecast Model Using Supervised Learning Algorithm,” IEEE Trans. Geosci. Remote Sens., DOI: 10.1109/TGRS.2021.3054582.指導教授 陳映濃(Ying-Nong Chen) 審核日期 2023-7-27 推文 plurk
funp
live
udn
HD
myshare
netvibes
friend
youpush
delicious
baidu
網路書籤 Google bookmarks
del.icio.us
hemidemi
myshare