請用此 Handle URI 來引用此文件:
http://tdr.lib.ntu.edu.tw/jspui/handle/123456789/86961
標題: | 生成式對抗網路用於多時序衛星影像除雲 CTGAN:Cloud Transformer Generative Adversarial Network |
作者: | 黃繼綸 Gi-Luen Huang |
指導教授: | 吳沛遠 Pei-Yuan Wu |
關鍵字: | 多時序衛星影像除雲,生成式對抗網路,哨兵衛星二號,福爾摩沙衛星二號,轉換器, Cloud removal for multi-temporal cloudy images,generative adversarial network,conformer,Sentinel-2 satellite,FormoSat-2 satellite, |
出版年 : | 2023 |
學位: | 碩士 |
摘要: | 雲的遮蔽干擾了遙測衛星影像的相關應用,包含環境監測、土地覆蓋分類以及貧窮預測等等。在這篇論文中,我們提出了Cloud Transformer Generative Adversarial Network (CTGAN)模型來進行衛星影像除雲,此模型的輸入為三張時序的有雲影像,輸出則生成一張對應的無雲影像。相比於過去的生成模型相關的文獻,我們特別著重在設計特徵擷取器 (feature extractor)使其能保留影像中無雲區域的權重,並且同時減少有雲區域的權重,接著把取出的特徵經過conformer,利用attention的特性,使其能在時序性的影像中找到最關鍵的特徵表示來進行還原。同時,為了解決在這個領域中少量資料的問題,我們自行從哨兵二號衛星蒐集資料進行資料標註,並且提出了Sen2_MTC資料集貢獻於這領域。最後,我們也在不同的衛星中進行廣泛的實驗,包括台灣的福爾摩沙衛星二號 (FormoSat-2)及歐洲的哨兵二號 (Sentinel-2),實驗表明我們提出的CTGAN不僅在這些資料集中都能達到state-of-the-art的結果,也在下游任務土地覆蓋分類中,有著顯著的進步,此論文的程式碼公開於此網址中https://github.com/come880412/CTGAN Cloud occlusions obstruct some applications of remote sensing imagery, such as environment monitoring, land cover classification, and poverty prediction. In this paper, we propose the Cloud Transformer Generative Adversarial Network (CTGAN), taking three temporal cloudy images as input and generating a corresponding cloud-free image. Unlike previous work using generative networks, we design the feature extractor to maintain the cloudless region's weight while reducing the cloudy region's weight. We then pass the extracted features to a conformer module to find the most critical representations. Meanwhile, to address the lack of datasets, we collected a new dataset named Sen2_MTC from the Sentinel-2 satellite and manually labeled each cloudy and cloud-free image. Finally, we conducted extensive experiments on FS-2, the STGAN dataset, and Sen2_MTC. Our proposed CTGAN demonstrates higher qualitative and quantitative performance than the previous work and achieves state-of-the-art performance on these three datasets. We also perform land-cover classification, which can be viewed as a downstream task after cloud removal. The improved performance on the land-cover classification demonstrates that our model has a high quality of generating cloud-free images compared to the previous works. The code is available at https://github.com/come880412/CTGAN |
URI: | http://tdr.lib.ntu.edu.tw/jspui/handle/123456789/86961 |
DOI: | 10.6342/NTU202300030 |
全文授權: | 同意授權(全球公開) |
顯示於系所單位: | 電信工程學研究所 |
文件中的檔案:
檔案 | 大小 | 格式 | |
---|---|---|---|
ntu-111-1.pdf | 678.67 kB | 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。