SAR影像的都市水體判釋：考量都市地表形態造成的雷達二次反射效果

Detecting urban water bodies from SAR images: Measuring urban surface morphology contributing to radar double bounce effect

廖晧宇(Hao-Yu Liao)

指導教授 ： 溫在弘

國立臺灣大學/理學院/地理環境資源學研究所/碩士(2019年)

https://doi.org/10.6342/NTU201900678

摘要

判釋水體在遙測研究中是個重要的研究議題。近年來，面對極端氣候的浪潮，歷時短、強度高的降雨越趨頻繁，使得地表水體的分布也變得更加多變與複雜。雖然透過光學影像已發展出數項水指標如NDWI、mNDWI來判釋地表水體，雲霧遮蔽卻為光學影像於應用上的最大限制。合成孔徑雷達影像（簡稱SAR影像）因其較長的雷達波長不僅能夠繞射雲層，同時，因水體表面的雷達回波特徵鮮明，SAR影像彌補了光學影像的天候限制，成為水體判釋常見的影像選擇。然而，斜視的主動式雷達於都市中的二次反射效果使得都市水體判釋面臨了挑戰。過去都市水體研究所採用的分析方法過度簡化了二次反射於不同地表形態及特徵下的作用效果，同時，在學理上也仍缺乏針對都市地區的水體判釋架構。本研究目的為，在二次反射的影響效果下，整合地表形態特徵並提出都市水體判釋的分析架構。其中，此研究所使用的資料包含Sentinel-1 SAR影像、都市建物圖資、數值地表模型（DSM）及數值地形模型（DEM）。方法上，本研究使用了羅吉斯回歸、支援向量機模式（SVM）及隨機森林（RF）。透過羅吉斯回歸模式釐清二次反射效果與地表形態特徵的關係，另一方面，使用機器學習模式以達到更好的水體判釋表現。本研究結果歸納了特定地表形態特徵在水體判釋上的二次反射效果，發現周圍水平方向密度以及特定高度（9公尺）的水平密度的重要性，並呈現二次反射效果在都市水體判釋中的空間分布。同時，也將分析架構套用於另一座城市（台中市）進行外部驗證，檢驗模式於跨城市應用的通用性。其中，以台中市的樣本資料驗證於台北市訓練出的羅吉斯回歸模式，達到0.90的判釋準確率。本研究考慮了SAR的二次反射效果提出都市水體判釋的分析架構，從中釐清都市的地表形態特徵於SAR影像判釋水體的重要性與異質性，並回應SAR影像中，對於都市水體的二次反射及判釋上的研究空缺。

關鍵字

雷達遙測 ； 合成孔徑雷達影像（SAR影像） ； 水體判釋 ； 都市表面形態 ； 二次反射效果 ； 水平密度 ； 平均高度

並列摘要

Extracting water extents and monitoring the dynamics of water bodies have been popular topics in remote sensing studies. In addition, under extreme weather conditions, the occurrence of surface water bodies can be much more dynamic due to floods or intense rains. While water bodies can be detected with water indices such as the Normalized Difference Water Index (NDWI) or the Modified Normalized Difference Water Index (MNDWI), cloud obscuration has limited the usage of optical images. Synthetic aperture radar (SAR) images validate the model generics can overcome weather conditions, can supplement the limited optical satellite images and have been frequently used for water detection. However, specific issues arise in urban areas due to the double-bounce effect. Previous studies have oversimplified the effect without considering the urban morphologies shaped by buildings and other structures on the ground, and there is no systematic urban water-detection framework. This study proposes a modeling framework for urban water-body detection and improves the prediction performances after including surface morphology variables in the models. The data include Sentinel-1 SAR images, building layers, digital surface models (DSMs) and digital elevation models (DEMs). Logistic regression, support vector machine (SVM) and random forest (RF) models are employed and compared. This study provides some implications for the mechanism behind the double-bounce effect, including the importance of the neighboring horizontal density and the horizontal density over a critical height (9 meters), and explores the spatial distribution of the double-bounce effect in urban water detection. Additionally, the models are tested in another city (Taichung City) to validate the models’ generalizability. The performances are satisfactory, with an accuracy of 0.90 for the logistic regression model that had been trained with Taipei City data and tested on Taichung City data. This study not only proposes an urban water-body extraction framework that considers the double-bounce effect but also illustrates the essentiality and heterogeneity of the neighboring surface morphologies when using SAR images for water extraction. Most importantly, the results of this study respond to the current research gap in urban water detection regarding radar double bounce.

並列關鍵字

radar remote sensing ； synthetic aperture radar (SAR) images ； water detection ； urban surface morphology ； double-bounce effect ； horizontal density ； mean height

參考文獻

Amitrano, D., Di Martino, G., Iodice, A., Riccio, D., & Ruello, G. (2018). Unsupervised Rapid Flood Mapping Using Sentinel-1 GRD SAR Images. IEEE transactions on geoscience and remote sensing, 56(6), 3290-3299.

Asner, G. P. (2001). Cloud cover in Landsat observations of the Brazilian Amazon. International Journal of Remote Sensing, 22(18), 3855-3862.

Bazi, Y., Bruzzone, L., & Melgani, F. (2005). An unsupervised approach based on the generalized Gaussian model to automatic change detection in multitemporal SAR images. IEEE transactions on geoscience and remote sensing, 43(4), 874-887.

Beres, M., & Haeni, F. P. (1991). Application of Ground-Penetrating-Radar Methods in Hydrogeologie Studies. Groundwater, 29(3), 375-386. doi:doi:10.1111/j.1745-6584.1991.tb00528.x

Bioresita, F., Puissant, A., Stumpf, A., & Malet, J.-P. (2018). A Method for Automatic and Rapid Mapping of Water Surfaces from Sentinel-1 Imagery. Remote Sensing, 10(2), 217.