以自相似算法進行衛星影像融合和水線判釋;Self-similarity algorithm for satellite image fusion and waterline interpretation

NCUIR > Research Center > Master of Science Program in Remote Sensing Science and Technology > Electronic Thesis & Dissertation > Item 987654321/90038

Please use this identifier to cite or link to this item: http://ir.lib.ncu.edu.tw/handle/987654321/90038

Title:	以自相似算法進行衛星影像融合和水線判釋;Self-similarity algorithm for satellite image fusion and waterline interpretation
Authors:	王家翔;Wang, Chia-Hsiang
Contributors:	遙測科技碩士學位學程
Keywords:	水線;凸自相似性正規化;影像融合;衛星遙測;waterline;convex self-similarity regularization;panchromatic sharpening;remote sensing
Date:	2022-09-28
Issue Date:	2022-10-04 12:08:49 (UTC+8)
Publisher:	國立中央大學
Abstract:	隨著多重感測器應用在遙感探測、電腦視覺等諸多領域的普及，多重感測器產品的融合影像儼然成為新興的話題。主要原因之一是各種感測器可以在同位置提供不同的時空影像。因此，本研究目標為合成來自不同感測器的全色銳化影像，並研究融合影像在水線檢測中的表現。該工作流程以具有低空間解析度但高時間解析度的Sentinel-2衛星影像為例，通過將該數據與全色態影像提供高空間解析度的SPOT-6 衛星影像進行融合。首先，我們將SPOT-6的全色態影像與Sentinel-2的多光譜 (NIR-B-G) 影像進行融合，使用自相似正規化全色銳化 (SimiRegPS)方法融合桃園地區的衛星影像，此自相似性已在自然影像以及各種成像逆問題中得到廣泛的驗證。然後，計算常態化差異水體指數全色銳化 (NDWIP) 以識別水像素。我們使用桃園市政府水務局提供的UAV正射影像驗證了場景一和場景二的8口埤塘，驗證包括旱季（場景一）和雨季（場景二）等兩種場景設定。在場景一中，融合影像中水線的平均精度在2.99 m和8.05 m 之間。在場景二中，融合影像中水線的平均精度在2.68 m和7.52 m之間。在場景一中，融合影像中水域的平均準確率為85%，而原始影像為 73%。在場景二中，融合影像中水域的平均準確率為84%，而原始影像為72%。綜上所述，本研究顯示通過將 Sentinel-2 與有限的 SPOT-6影像相結合，通過SimiRegPS方法獲得更準確的水線，可以有效地提取水文參數。;With the popularization of multi-sensor applications in remote sensing, computer vision, and many other fields, the fusion of multi-sensor products has become an emerging topic in the community. One of main reasons is the variety of sensors can provide different spatiotemporal images in the same location. Hence, this study aims to compose a panchromatic-sharpened image from heterogenous sensors, and to investigate the performance of the fused image in waterline detection. The workflow is exemplified by Sentinel-2 that has a lower spatial but high temporal resolution, and to merge the data with SPOT-6 that provide much higher spatial resolution in its panchromatic band. We first fuse the panchromatic images of SPOT-6 with the multispectral (NIR-B-G) images of Sentinel-2, by using the Self-similarity Regularized Pansharpening (SimiRegPS) method to fuse the images covering Taoyuan, Taiwan. The self-similarity employed in our design has been extensively examined in natural images as well as in various imaging inverse problems. Following that, the Normalized Difference Water Index Pansharpened (NDWIP) is calculated to identify water pixels. We validate 8 ponds as compared with in situ data from Taoyuan Water Resources Department. The validation includes two scenarios: dry season (scenario 1) and wet season (scenario 2). In scenario 1, the averaged accuracy of waterline in the fused image is between 2.99 m and 8.05 m. In scenario 2, the averaged accuracy of waterline in the fused image is between 2.68 m and 7.52 m. Also, the averaged accuracy of water area in the fused image is 85% and 84%, in contrast to 73% and 72% of the original image in scenario 1 and 2, respectively. To conclude, this research has shown the possibility to effectively extract hydrologic parameters by combining Sentinel-2 with limited SPOT-6 images to obtain the more accurate waterline through SimiRegPS method.
Appears in Collections:	[Master of Science Program in Remote Sensing Science and Technology ] Electronic Thesis & Dissertation

Files in This Item:

File	Description	Size	Format
index.html		0Kb	HTML	81	View/Open

社群 sharing

Loading...