衛星影像灘線辨識之精確度評估研究

Abstract

本研究主要提出能精確擷取衛星影像中灘線之方法，其包括提高衛星影像圖解析度的HSI影像融合法，合併新定義的NDWI參數於加密的HSI值，再以最大概似法進行監督式分類之。原本所擷取衛星影像圖中的灘線與實測灘線之誤差RMSE約為7m。因不同拍攝時間的衛星圖潮位不同，對於潮位影響平均灘線位置，本研究提出以實際坡度與經驗坡度公式兩個修正方法，將原灘線修正至平均潮位基準之灘線，經二者方法的得修正灘線與實測值之誤差RMSE降低為6m左右。本研究結果說明若沒有實測坡度可作為修正依據時，藉由現地的波浪特性與底質粒徑所得之經驗坡度，來修正灘線位置是可行的。 最後本研究以此灘線辨識方法，擷取台南的黃金海岸2004年至2008年的灘線，由分析灘線變化結果顯示，黃金海岸為一個夏侵冬淤的海岸，尤其南端灘線在2004年與2005年兩年間曾發生劇烈的侵蝕現象，整體平均後退了100m，北端則是沒有明顯的進退，處在一個平衡狀態，此結果與往者研究之南侵北淤略有不同。本研究採用西班牙大學研發出來的海岸模擬系統(SMC)，來模擬黃金海岸冬夏季的漂沙情況，計算結果顯示夏季的漂沙趨勢為將侵蝕近灘岸之沙帶到外海堆積，而冬季的漂沙為僅微量的岸上之漂沙被帶至外海，此結果說明黃金海岸以向離岸為主的漂沙特性，且呈現全年侵蝕的狀態之原因。

A method for accurately extracting shorelines of satellite images is investigated in this study. Above all the pixel resolution of a satellite image is doubled using fused HIS (Hue, Saturation, Intensity). The shoreline of a satellite image is detected by supervised classification on fused HIS associated with a newly-defined NDWI using maximum likelihood method. The detected shoreline of a chosen satellite image is compared with the measured shoreline in site to have a RMSE (root mean squared error) of about 7 m. Due to different tidal levels of different satellite images the detected shorelines are not based on the same datum. The originally detected shorelines can be shifted to the mean water level using tidal deviation from the mean water level associated with measured or computed bottom slopes. The corrected shoreline is closer to the measured shorelines than the original one by a RMSE of about 6 m. The computed bottom slope from an empirical formula including wave height and period and sand diameter is applicable for shifting the detected shoreline to the mean water level when the measured bottom slope is unavailable. The examined shoreline detection of a satellite image is applied to investigate the shoreline revolution of Gold beach in Tainan from 2004 to 2008. The change of detected shorelines at the same datum shows that the beach erodes in the summer and deposits in the winter. An average of shoreline retreat about in the south part from 2004 to 2005 is demonstrated to be about 100 m but the beach in the north part is in an equilibrium state. The result is different from the previous work indicating erosion happens in the south part but deposits in the north part. The reason for the beach sediment is explained using the computed sediment on the beach by SMC (coastal model system developed by Ocean and coastal research group of university of Cantabria). Offshore sediment is dominant so that the beach is eroded by large waves in the summer. The small waves in the winter have little impact on beach change. On-offshore sediment is proved to erode the Gold beach by the present study.