Improvement of DEM generation from ASTER images using satellite jitter estimation and open source implementation |
Paper ID : 1112-SMPR-FULL |
Authors: |
Luc Girod *1, Christopher Nuth2, Andreas Kääb2 1Prestegårdsveien 53 2University of Oslo |
Abstract: |
The Terra (EOS AM-1) satellite was launched in December 1999 on a Sun-synchronous orbit. Aboard this satellite is the ASTER (Advanced Spaceborne Thermal Emission and Reflection Radiometer) system. For more than 15 years now, sets of stereo images were collected by ASTER all over the world at a 15m resolution in the near infra-red band, making its data the largest consistent multi-temporal worldwide dataset of stereo images available. The ICEMASS project aims at analyzing glacier thickness changes using satellite data at a global level. The incredible goldmine of data produced by ASTER can be used for 2D/3D movement estimation and change assessment in glaciological research, providing several consistent snapshots through a time series. However, DEMs generated by NASA with SilcAst [Sensor Information Laboratory Corp, 2004] (ASTER DMO products) do not provide a sufficient geometric quality for glacier volume change estimation over short periods, the expected change being significantly smaller that the accuracy of the product (a few meters against +-30m). This is due in part to high frequency satellite jitter caused by unknown sources. The jitter yield at least three superimposed sinusoidal signals in DEM comparison between the ASTER DMO and ground truth [Nuth and Kääb, 2011] or between ASTER and QuickBird images [Ayoub et al., 2008].The low acquisition frequency of the GNSS/IMU platform and star camera prevents the on-board estimation of jitter using the meta-data alone. A solution was proposed by [Iwasaki, 2011] to correct the jitter in the band 3N (pseudo-nadir, near infra-red band) using the very short time delay with the other pseudo-nadir bands (1 and 2). However, it cannot be applied to correct the band 3B (back-looking, near infra-red band) since only one band is captured by the back-looking telescope. We aim to identify and correct jitter for the band 3B as this has the largest effects on ASTER DEM quality. Our method evaluates the inaccuracies in co-registration due to imprecise orbital and orientation data of the Level 1A ASTER products without the use of ground control points (GCP). The use of GCPs is not realistic for a method designed to process ASTER images globally. It starts by estimating a RPC model (Rational Polynomial Coefficient) for both images of the stereo pair (bands 3N and 3B) using seed points taken at different altitudes from the lines between the satellite and the lattice points, a method similar to the one described in [Lutes and Grodecki, 2008]. From these RPCs, we then generate relationship grids between image and world coordinates. The relative offset between the two images and the exact parameters of the jitter (frequencies, phase and variation in amplitude) of the back looking image are then estimated using tie points. These estimates offer an improved relative accuracy of the DEM compared to ASTER DMO products without the use of GCPs. Our implementation of the estimation and correction method is in the open source MicMac photogrammetric library (developed at the French Geographical Institute – IGN – [Pierrot-Deseilligny et al., 2015a] and [Pierrot-Deseilligny et al., 2015b]), hence offering a ready to use software to produce DEMs and ortho-images from Level 1A ASTER products. These improvements have the potential to open the ASTER DEM archive for Global terrain change analysis where changes are often smaller than the pixel sizes, as is often the case for glacier elevation changes. |
Keywords: |
ASTER, JITTER, DEM, GLACIER, CHANGE |
Status : Paper Accepted (Oral Presentation) |