近年來，遙測影像的應用範圍愈來愈廣泛，在製作地圖方面經常會使用到，甚至在工程上也有使用，但是，在使用遙測影像之前還需要經過一些處理，其中最重要的就是幾何校正這方面，一般來說，影像校正是使用載具上所記錄的飛行軌道資訊來進行，但是若無軌道資訊時，就會取一張level-4影像，在兩張影像相同地點上選取校正點。而不同載具所使用的光譜波段不盡相同，不同光譜的反射量值對相同物體也不一樣，不過在現行的自動選取校正點方法中，通常是使用相關係數來選出同樣的地區，再從這裡找出特徵點，可是這方法在兩張不同載具所拍攝的影像上就無法使用，原因在於相關係數是使用反射量值來做選取，光譜不同的話，找出來的區域就很可能不符，所以這份研究是希望找出一個方法，讓不同來源的影像可以找出相同位置的校正點，藉此減少人工找點的時間。 此論文使用 ”完全限制最小平方差” (Fully Constrained Least Square) 分類法來萃取出地面上的資訊，如道路和湖泊，再從這些資訊中，找出特徵點做為校正點，如十字路口和湖泊幾何中心。Remote sensing image become more and more popular in near years. It also is used in engineering and making map. But the image can not use directly when we get the raw image data (level-1 image). The main method to solve this problem is using satellite’s flight information to do geometry correction. If we do not have this information, we can take one image which was corrected (level-4 image) to decide the tie points. Generally speaking, the tie points automatic extracting is using correlation in two different level images to find same feature being tie points. Correlation method is using reflectivity to select same area. It can not use correlation to decide the same position, because the reflectance of one material would different in each band. And the wave bands which different satellites used are not always being same bands. In this research, we want to find a method to decide the tie points in different sensor images. We use the Fully Constrained Least Square (FCLS) classification to extract the information on ground. Our opinion is using the information on the ground to find the same feature, like roadd and lakes. Then detecting cross road and barycenter of lake to be tie points. We hope this method can help operator find the tie points easier.
