ESA       
   workshops       
envmail@esa.int

 

DSM Generation Using ALSO/PRISM Image Data and the software package SAT-PP

Kirsten Wolff(1) and Armin Gruen(1)

(1) IGP, ETH Zurich, Wolfgang-Pauli-Strasse 15, 8093 Z├╝rich, Switzerland

Abstract

One of the most important products of ALOS/PRISM image data are accurate DSMs. To exploit the full potential of the high resolution of the sensor for the DSM generation, among other things, like a sophisticated camera model, powerful software packages for the execution of various other functions are needed. Our group has developed a suite of new methods and a software package called SAT-PP (Satellite Image Precision Processing), which can be used for the 3D processing of all optical linear array CCD-based satellite sensor systems. The main module of the software is the multiple images matching process for DSM generation. During the generation process we use relational polynomial coefficients (RPCs) for the projection between object and image space. These coefficients can either be imported or can be derived from the parameters of our rigorous sensor model. The image matching process and the final DSM generation are nearly fully automatic. The only manual step is the measurement of some seed points in the stereo model to get a simple approximation for the difficult parts of the DSM if no other initial DSM is available. As a member of the validation and calibration team for ALOS/PRISM, we published first results for calibration, georeferencing and DSM generation using our software package SAT-PP. For this validation we defined areas with different characteristics like open, forest, alpine and city areas and determined the root mean square errors from reference data of these areas. Although no additional manual post processing was done, the results show the high potential of our software package in general and also the potential of the image data of ALOS/PRISM. Subpixel accuracy is possible for well defined features and the overall accuracy across all object and image features lies between 1-5 pixels, depending primarily on surface roughness, vegetation, image texture and image quality. Here we will discuss our newest results of the quality control and evaluation of ALOS/PRISM in combination with our software. We will focus on local critical areas, like steep hillsides. To improve such areas, additional seed points can be easily measured in the stereo mode. By this, the globally high quality DSM can be improved locally. Beside this, different other aspects of a detailed quality control are addressed, like the potential of the sensor using three views simultaneously in comparison to just using only two.

 

 

  Higher level                 Last modified: 07.05.06