頻譜分析應用於決定衛星影像中的波向及水深推估

Abstract

往昔水深資料的獲得方式皆為利用船隻逐點進行量測，雖船隻上裝有GPS定位系統，但對於實際量測路線規劃上仍然無法精確掌握。故本文將快速傅立葉轉換的特性運用於衛星影像圖中，經由頻譜圖決定波向並推估水深。 本文選擇龜山島海域及屏東縣枋寮漁港外側海域作為研究基地，龜山島海域海岸線曲折且水深變化大，較能看出波浪特性；而枋寮海域海岸線平直無曲折雖波浪特性不明顯但有實測水深數據可供模式推算驗證。 本文首先運用直方圖均化提高影像對比度再採用Butter Worth高通濾波濾除雜訊，最後由頻譜圖求得波向及波長。考慮波浪平緩前進至近岸的情況下利用波浪折射理論修正波長，並運用波浪理論中的分散關係式估算水深。 本文利用屏東海岸實測水深資料評估本模式推算水深之精確度，本模式估算於-4m至-10m的水深較實測水深高，而-10m至-12m間兩者接近，-12m以下的水深則略低於實測水深，估算水深與實測水深的均方根誤差約為0.49m，若不考慮波紋線較凌亂的部份其均方根誤差則降低至0.31m，結果顯示本方法在估算水深具有一定的可信度，未來可應用於港灣建設的事前規劃。

In the past, the way to get bathymetry data was using vessels measuring point by point although the vessel was equipped with a GPS positioning system. However, route planning on the actual measurements was still unable to grasp accurately; therefore, this paper is going to illustrate how Fast Fourier Transform on the satellite images can determine wave rays and bathymetry. This paper choose Guishan Island and Pingtung Fangliao coast as research base. The Guishan Island coast water changed significantly, which makes it clearer to observe the wave characteristics. On the other hand, Fanliao Sea coastline is straight without twists. The wave characteristics are not obvious, but there are bathymetric data available for verification. Some techniques of image processing are used to obtain wave rays and water depth. The techniques include: (1) histogram equalization for increasing image contrast; (2) Butter Worth high-pass filter for filtering out the noise; (3) Determining wave rays and wavelength by using spectrum. Finally, consider waves passed to the shore gently, using theory of wave refraction correction wavelength and estimate bathymetry. The proposed method is assessed valid for estimating water depth by comparing the results obtained by bathymetry survey with those obtained by the proposed method at the Pingtung coast. The estimated water depths by the proposed methods are from -4m to -10m, which is higher than the data by the bathymetry survey. However, for deeper water depth exceeds -12m, the estimated water depths by the proposed methods are slightly lower than those by the bathymetry survey. Both results are approximate between the water depth from -10m to -12m. The root mean square error between the estimated water depths and measured ones along a profile is 0.62m. If we don’t consider the part of messy wavy lines, its root mean square error is reduced to 0.31m. The results indicate that the method of estimating the water depth has certain credibility. In the future, this method can be used to harbor construction planning.