DEM generation and accuracy analysis from high resolution satellite images through a 3D correlation matching algorithm

초록

Research of accurate DEM generation from high resolution satellite images has been an active research topic for many years. To overcome problems related to automatic stereo matching of such images, we have proposed a new correlation matching algorithm (3DCOM) that works in a 3D object space. In this paper, we aim to test the 3DCOM to various high resolution satellite images and discuss it performance. We applied the 3DCOM to Worldview-1, IKONOS and Kompsat-2 stereo images. The accuracy of the DEM was analysed against the ground truth data at 1m grid spacing acquired from LIDAR. The root mean square error of the DEM generated from the Worldview pair was less than 3.5m. Our experiments support the successful development of the 3DCOM and the merit of object space matching in handling high resolution satellite images