Analysis of ALOS-PRISM image geometry for advanced photogrammetric applications

This study was launched at IGN-France to better understand ALOS-PRISM image geometry and be able to use these images for photogrammetric applications especially DSM generation. From provided ancillary data, we managed to build PRISM rigorous location model with location accuracy around 70m without GCP and 2.5m after adjustment by a simple bias correction of pitch, roll, yaw and focal length with GCPs from Reference3D data. Dense image matching and statistical analysis of shifts between triplet images highlight a very accurate attitude determination but also remaining errors of focal planes calibration (up to 1 pixel). A DSM generated from a PRISM triplet over Rio, Brazil reaches DTED2 standard with topographical details like ridges and valleys well rendered. But comparison with Reference3D DSM also shows defects due to these remaining errors of focal planes calibration.