近年來航海技術日趨發達，海上運輸及進出口貿易的次數也日漸頻繁，在這樣的情況下，海洋溢油汙染事件發生的機率也大幅提升，而這會對海洋生態環境造成嚴重的衝擊，因此如何偵測、監控並追蹤溢油汙染變成一項重要的議題。但由於在寬廣的海洋上，不易取得現地量測的資料，在這樣的情況下，能夠穩定且簡單獲取廣大地表資訊的遙測影像，提供海洋溢油汙染偵測良好的資訊。 在本次研究中，我們提出一套利用光學多光譜衛星影像來進行油汙偵測的技術。通常海洋溢油汙染分布只佔了衛星影像中的極小範圍，且油汙與海水的光譜特性有極大的不同，根據這樣的特性，我們將海洋油汙視為異常物質，並且針對影像進行異常物偵測(anomaly detection)。在執行異常物質偵測之後，可能為油汙的區域會明顯地顯示出來，但此偵測到的區域也同時包含了由海面波浪所造成的雜訊，為了要降低由海浪造成的誤差，依據油汙與波浪在影像中的分布型態明顯不同，我們計算影像的空間特徵資訊 (spatial feature information) 並加入流程之中來提升偵測結果。同時我們也引進了數學型態學 (mathematical morphology) 的概念，來更進一步地移除偵測結果中的雜訊。 我們所提出的偵測方法同時使用了影像的光譜資訊及空間特徵資訊，並且結合數學型態學，藉由這樣的方法，海洋油汙的分布能被正確地偵測出來。 Oil spill on the sea surface which is usually produced by human activities is disastrous to the ecological environment. In recent years, the technology for sea transportation is improvement. The export and import via marine transit are more and more frequently, the oil spill events are sometimes along with the process. How to detect, monitor and track the oil spill are always very important tasks. Due to oil spill often occur in open sea, remotely sensed image provides an effective technology to monitor the sea area. But the oil spill usually only occur in a very small area in the image. How to detect the oil spill on the sea surface is a challenge problem. Using the remotely sensed images to detect this unwelcomed hazard material on sea surface is a convenient and effective approach. This study focuses on the detection of oil slick on sea surface using multi-spectral imagery technology. Since the oil spill area is usually very small compare to the sea area in the image scene, it can be considered as anomaly. By using anomaly detection algorithm, the oil spill can approximately figure out, but there are still some marine phenomena to interference the result. In order to discriminate oil slick from the other interferences, spatial features are introduced into anomaly detection. Furthermore, we also adopt mathematical morphology to filter through the maintain noise further. Therefore, our proposed method is to employ the spatial features of oil spill and combine with spectral information and mathematically morphologic operation to improve the oil spill detection. In the experiment, we adopt SPOT multispectral images for performance analysis.
