在過去幾年，有許多的學者做了許多有多關河道擷取的研究。而他們多數都是利用數值地形模型(DEM)及地理資訊系統(GIS)去達到目標，但是這些技術會有資料沒有更新的問題，所以無法得到目前河道確切的位置，另一方面，衛星每天拍攝地表，所以我可以利用遙測影像來做處理及偵測河道當下正確的位置。 這份研究會利用以下的方法去做擷取，首先，我們利用局部性灰階等化來增強影像對比，並用方向性中值濾波去消除雜訊，接著利用影像分類法來標示河道。為了改善雜訊及連接細的分支，再利用形態學的方法來處理。最後，我們利用遮罩去除單一雜點、利用霍夫轉換去連接斷線，以及去除短線等作法。這個研究的技術也可以用來處理醫學影像，我們會用角膜影像來證明血管擷取方法。There are researchers devoted their effort on river detection in the past decades. Most of them use digital elevation model (DEM) or Geographic Information System (GIS) for this task, but this types of data is seldom up to date. Therefore, the detection result may not represent the location of the rivers at present. On the other hand, satellites collect the remote sensing images from the ground every day, so we can detect the river and analyze the changes with recently acquired images. The objective of this research is to develop an algorithm with image processing techniques to detect river automatically. Our proposed method includes the following steps. Firstly, adaptive histogram equalization method is applied to enhance our image, and then use direction median filter to reduce noises, sequentially classify image to locate rivers. We also use morphology method to remove noise and connect narrow breach. Finally, we use mask detect single-point and remove them, apply hough transform to connect broken line and delete short line deletion. Our method can also be applied on medical image. We will also demonstrate the detection of vessels in cornea image.
