DSM Generation from early ALOS/PRISM Data using SAT-PP

Abstract

The new Japanese optical sensor ALOS/PRISM which was launched in January 2006 has stereo capabilities which can be used for a great number of applications and products. One of the most important products is the generation of Digital Surface Models (DSMs). With SAT-PP (Satellite Image Precision Processing) our group has developed a suite of new methods and a software package which can be used for the 3D processing of optical Linear Array CCD – based satellite sensor systems. The software was already used successfully for the DSM generation with e.g. IKONOS, QuickBird and SPOT5 high resolution image data. Here we will demonstrate the functionality of the automated image matcher for the generation of DSMs by using early ALOS/PRISM data with a ground sampling distance of 2.5 meters. We have tested our methods of DSM generation using the ALOS/PRISM dataset of two different testfields, which are located in theareas of Bern/Thun (Switzerland) and Okazaki (Japan). They have different topography and land use features (mountainous, city, forest, open and mixed areas). For a detailed quality control of the DSM generation, reference data for three sub areas of the testfield Bern/Thun were available and one sub area in Okazaki. A detailed analysis is presented for certain areas with special and homogeneous topography or land use characteristics. The overall height RMSE for the three sub-areas was 2-3 pixels depending primarily on surface roughness, vegetation and image texture and image quality. The height RMSE values for the specific areas ranged from less than 2 pixels (open areas) to 5 pixels in a tree-covered area. ALOS/PRISM images have particular radiometric problems which do not influence the final results significantly, but they can influence the DSM generation locally. Therefore we give also an analysis of image artifacts.