利用FFT及BEMD方法決定衛星影像圖的波向線並推算海下水深的研究

Abstract

解決海岸侵淤之問題，從海岸地形變遷的特性來了解是基本的研究工 作。一般海灘及海底地形的測量範圍可能涵蓋沿岸10公里而離岸2-5公里 左右的大範圍，若以往昔利用船隻載具單束或多束測深儀來量測海底地形， 在大範圍的此種測量方法不僅費時、費力，而且測量經費相對會高，測量頻 率的要求為一年兩次，但一般在經費的限制之下，無法長期且密集的量測。 本計晝主要提出以解析度為2公尺的福衛2號衛星影像來決定波向線及推算 水深的新技術，以克服往昔大區域測量的問題，或補足工程規劃時缺少往昔 水深資料的參考。

本計劃擬以兩年時間來執行。第一年使用前期計晝的影像前置處理方法 如直方圖均化、空間濾波及Otsu演算法先擷取衛星影像圖中的波紋線。再 利用快速傅立葉轉換法(FFT)方法以波向線來決定各點的波向，再由擷取的 波長及波向透過線性波浪理論地分散關係來推算水深，選擇屏東及宜蘭海岸 當為驗證本研究方法可行的兩個基地。

因為現場風浪除主要波以外還有其他短周期的小波及因地形或結構物 的反射波所組合而成的，因此影像圖波紋會呈現雜亂現象。為解決此問題第 二年擬用二維經驗模態分解法(BEMD)來進行濾出衛星影像圖内主要波浪 的波紋，BEMD已被應用於不同的訊號處理。本年會撰寫三種不同演算法 的BEMD程式，來分解出主要波浪的波紋且評估何種演算法比較適合分解 出主要波浪的波紋。再利用第一年的研究基礎以主要波浪的波紋來推算水 深，並比較兩年的研究結果的正確性。本創新的研究在國外尚在啟蒙階段， 國内尚無人進行，本研究結果可以應用至海岸工程規劃所需的海底地形資料 的參考。

To know the characteristics of bathymetric evolution is a fundamental method for solving the problem of beach erosion. A large region of bathymetric measurement is generally considered as a nearshore cell with an area of 10x(2-5) km . The measurement of a vehicle with single/multi -beam echo sounder for such large region needs high cost and intensive labor. Intensively previous measurements are generally unavailable in coastal planning due to budget limit. In the project a new technique of estimating underwater bathymetry will be developed to solve the problem of lacking underwater bathymetry in coastal planning using satellite images.

The project is planned to be carried out for two years. Based on the image processing of the previous project, such as histogram equalization, spatial filtering, image erosion, and Otsu algorithm, Fast Fourier transformation is used to determine a wave ray and wave lengths from extracted wave crests of wave trains in satellite images in the first year. Underwater bathymetry can be estimated using the dispersion relation of Airy wave theory. Bathymetry of Peng-Tung and I-Lan coasts is chosen as evaluation of applicability of the proposed method.

Some capillary waves by wind or reflected waves due to marine structures on a primary wave make the complexity of wave pattern in the real sea. Excluding capillary waves, primary wave should be determined for accurate estimation of wave length. In the second year, IMF (instinct frequency mode) of two-dimensional empirical mode decomposition (BEMD) method is used to indicate the primary wave of satellite images. BEMD is widely applied to different signal processing. Programs of BEMD are developed for three algorithms in this year. Through detected primary waves associated with the techniques developed in the first year model accuracy will be evaluated by comparing with the previous results. Estimation of bathymetry from satellite images is the beginnings of image processing in the foreign researches, but is absent in the domestic research. Results from this innovative research can provide with a reference to bathymetry in coastal planning.