Skip navigation

DSpace JSPUI

DSpace preserves and enables easy and open access to all types of digital content including text, images, moving images, mpegs and data sets

Learn More
DSpace logo
English
中文
  • Browse
    • Communities
      & Collections
    • Publication Year
    • Author
    • Title
    • Subject
    • Advisor
  • Search TDR
  • Rights Q&A
    • My Page
    • Receive email
      updates
    • Edit Profile
  1. NTU Theses and Dissertations Repository
  2. 工學院
  3. 土木工程學系
Please use this identifier to cite or link to this item: http://tdr.lib.ntu.edu.tw/jspui/handle/123456789/97799
Title: 基於資料立方架構之光學衛星影像雲偵測與除雲技術
Cloud Detection and Removal Techniques for Optical Satellite Imagery Based on a Datacube Framework
Authors: 江冠均
Guan-Jyun Jiang
Advisor: 韓仁毓
Jen-Yu Han
Keyword: 雲層偵測,雲層去除,深度學習,泊松融合,資料立方,
Cloud Detection,Cloud Removal,Deep Learning,Poisson Blending,Datacube,
Publication Year : 2025
Degree: 碩士
Abstract: 本研究針對光學衛星影像中雲層遮蔽問題,建構一套結合深度學習、資料立方與自動化影像融合之雲層偵測與除雲系統。首先,於雲層偵測階段,採用五種語義分割模型進行性能比較,結果顯示基於 Transformer 架構的 CSDFormer 於雲區識別表現最為優異,達成 91.08% 的雲區準確率(PA)、93.78% 的使用者準確率(UA)與 97.25% 的整體準確率(OA),並具備最快的推論速度與最低計算複雜度。其次,於除雲流程中,本研究以 SSIM(結構相似性指標)為排序依據,結合泊松融合進行雲區補全,於高達 80% 雲覆蓋情境下仍能維持 0.7105 的 SSIM,顯示良好的視覺品質與輻射一致性。最後,透過自行開發之資料立方系統整合影像屬性、空間範圍與坐標資訊,實作多條件篩選、影像排序與動態除雲操作,前端介面以 Flask 架構實現互動式展示,強化系統應用實用性。整體成果顯示,本研究提出之架構於多時序光學影像管理與除雲應用中具備高準確性、擴充性與操作效率,未來可作為光學遙測資料品質提升與資料治理整合之基礎平台。
This study proposes an integrated system for cloud detection and removal in optical satellite imagery, combining deep learning, a Datacube framework, and automated image fusion. Among five evaluated segmentation models, the Transformer-based CSDFormer achieved the best performance (PA: 91.08%, UA: 93.78%, OA: 97.25%) with the fastest inference speed and lowest complexity. For cloud removal, SSIM-based image ranking and Poisson blending were employed, maintaining an SSIM of 0.7105 even under 80% cloud cover. A custom Datacube was developed to manage metadata and spatial queries, supporting dynamic filtering and processing. The system, implemented with a Flask-based interface, shows strong accuracy, efficiency, and scalability for enhancing multi-temporal optical imagery.
URI: http://tdr.lib.ntu.edu.tw/jspui/handle/123456789/97799
DOI: 10.6342/NTU202500924
Fulltext Rights: 同意授權(限校園內公開)
metadata.dc.date.embargo-lift: 2025-07-17
Appears in Collections:土木工程學系

Files in This Item:
File SizeFormat 
ntu-113-2.pdf
Access limited in NTU ip range
17.64 MBAdobe PDF
Show full item record


Items in DSpace are protected by copyright, with all rights reserved, unless otherwise indicated.

社群連結
聯絡資訊
10617臺北市大安區羅斯福路四段1號
No.1 Sec.4, Roosevelt Rd., Taipei, Taiwan, R.O.C. 106
Tel: (02)33662353
Email: ntuetds@ntu.edu.tw
意見箱
相關連結
館藏目錄
國內圖書館整合查詢 MetaCat
臺大學術典藏 NTU Scholars
臺大圖書館數位典藏館
本站聲明
© NTU Library All Rights Reserved