近岸水深海圖測繪是海洋國家重要的基礎圖資，也是許多研究和應用必要的基本資料。傳統以 海測船利用聲納進行水底探測成本非常高，而且很沒效率，也難以建立全面且準確的海底地 形。利用主動或被動式遙測資料進行水深反演雖然精度比不上聲納探測，也無法完全取代傳統 的船舶聲納測量，但是在船測不易進行的近岸淺水區是一個可行的水深探測替代方法。本計畫 以研議三年為期，發展線性及非線性多光譜水深反演模式，以利用WorldView-2高解析多光譜 衛星遙測影像進行南海研究區內的島礁近岸水深反演與海底地形模型建置。此外，由於 WorldView-2影像的頻寬較大，增加模式參數推導的不確定性，因此所發展的線性與非線性光 譜水深反演模式將結合粒子群最佳化演算法，進行水深反演與參數推導的優化，提高水深反 演與海底地形模型成果的可靠性。 ;Bathymetry maps of ocean islands are important for both safe navigation and scientific studies. These maps can be generated from different data sources and using different methods. In-water ship-base surveying, is one of the known methods known to produce the most accurate depth information but is limited to deep-water areas. High-resolution multispectral satellite imagery allows mapping of underwater features that are not accessible to surveying vessels. Many models have been developed to extract water depth from multispectral imagery. These models, although have been successfully applied in various applications, have been shown to perform poorly in optically shallow waters with heterogeneous bottom types and varying albedo. This project proposes to develop linear and nonlinear multi-spectral bathymetric models in order to utilize high-resolution WorldView-2 multi-spectral images to generate accurate and complete bathymetric maps of shallow water areas in the study size located in South China Sea. Because the bandwidth of WorldView-2 imagery is relatively wide, which will increase the uncertainty of the model parameters and derivation, the spectral-bathymetry models will also integrate with Particle Swarm Optimization (PSO) to increase the reliability of the derived bathymetric results.
