傳統上，衛星進行幾何校正時，必須選取控制點，通常選取過程是以人為的方式比對影像與地圖，將影像中屬於明顯的特徵選取出來，在地圖上找到對應的座標後，才能對影像做幾何校正。高解析力的衛星影像，因具有空間特徵物相當明顯的特性，更需要做幾何校正，故本研究應用數位影像處理的技巧簡化萃取控制點座標的過程，使高解析力衛星影像能夠迅速取得地面控制點。 數位影像處理的過程可分五個部分，一、道路特徵之萃取，以分離衛星影像中道路面的可能資訊為目的，產生二元特徵影像；二、細部雜訊之消除，以降低特徵影像中的雜訊並進行特徵區塊編碼為目的，產生特徵區塊影像；三、道路特徵區塊之辨識，以辨識特徵區塊影像中的道路特徵區塊為目的，產生道路特徵區塊影像；四、道路特徵區塊形態之修正，以修正道路特徵區塊形態提供符合人類認知的道路資訊為目的，產生道路二元影像；五、道路交會口交點之搜尋，以輸出道路交會口交點位置為衛星影像之一個控制點為目的，產生一組控制點座標。 透過上述的數位影像處理步驟，在挑選高解析力衛星影像控制點時，便能夠降低人工介入的程度，迅速、確實的自動化萃取高解析力衛星影像控制點座標。 The high spatial resolution satellite images have widely attracted the attention in the remote sensing community recently. It is obvious that one of the great challenges to process the high spatial resolution satellite images will be the geometric correction practice. Conventionally, the positioning of the image control points is manually performed by a labor-intensive and time-consuming procedure. The practicable experience indicates that the qualified image control points would be the points with striking features such as the road intersection. Thus, due to the abundant image contents, high spatial resolution satellite image would have plenty of the qualified control points. As a result, the manual identification and positioning of control points will become even more inefficient and unbearable. Therefore, the main objective of this study aims to develop an automated image processing technique to extract the control points for the high spatial resolution satellite images. Among numerous spatial features, this study considers road intersection the main target to perform the control point extraction. The proposed image-processing algorithm consists of three steps. The first step is designed to segment the image and produce the feature image. The second step is proposed to extract roadblock features from the feature image. The third step is planned to locate the center position of the roadblock. A series of high spatial resolution satellite images are used to test the proposed method. The preliminary results shows that the proposed image processing approach has the potential to automatically position the control points in the high spatial resolution satellite image.
