Geometric Potentiality of GoeEye-1 In-Track Stereo Pairs and Accuracy Assessment of Generated Digital Surface Models

The Geoeye-1 satellite, launched in September 2008, is able to acquire imagery in panchromatic mode and in multispectral mode with a spatial resolution of 0.41 m and 2.0 m at nadir, being the highest resolution sensor from space presently commercially available.

The aim of the work is to evaluate the quality of the GeoEye-1 stereo pair panchromatic product through the accuracy assessment of the orientation and of the generated Digital Surface Model (DSM).

The area covered by the stereo pair is the city of Rome, a long time established test field where different products acquired by several sensors were checked during the past years. The analysed stereo pair has 0.5 m pixel size and it belongs to the GeoStereo product class; it was acquired on December 5, 2009, but it was made available only at the end of January 2010, so that, at present, only very preliminary analyses were carried out.

The GeoStereo products are radiometrically corrected map oriented image, suitable for a wide range of uses. GeoStereo images are projected onto an “inflated” ellipsoid, derived from the WGS84, choosing a certain ellipsoidal height.

Thanks to the very high resolution and the good radiometric quality of the images, it seems possible to extract a DSM at a 2-3 meter accuracy level, comparable to those coming from middle scale aerial products, supporting a wide range of geomatics applications such as mapping, spatial and temporal change detection, feature extraction and data visualization.

The stereo pair orientation has been carried out with the rigorous models embedded into the scientific software used is SISAR, developed at the Geodesy and Geomatic Area of the University of Rome ”La Sapienza”; it is based on well know collinearity equations, with the reconstruction of the orbital segment during the image acquisition and possibly self-calibration parameters estimation. The results have been checked with the commercial software PCI Geomatica 10.2, which includes the well known Toutin's rigorous model.

Further, the image matching has been performed with the new matching facility of the software SISAR and were checked both with PCI Geomatics v. 10.2, and with the scientific software DPCOR, kindly supplied by Prof. Jacobsen (Leibniz University of Hannover).

The new SISAR image matching facility is presently based on two different algorithms. In the first algorithm the advantages of ABM to those of FBM has been combined to decrease the efforts for the initial manual points collimation needed to start the matching process. The second one uses a two-stage dynamic programming technique to obtain a disparity map from a correlation volume; in this way it was possible to improve the reliability and accuracy of matching in urban area; this approach has been already tested on stereo pairs with quasi-epipolar geometry (like GeoEye-1), as Ikonos stereo products.

In order to evaluate the accuracy of the DSM extracted a comparison with 3D vector cartography at 1:2000 scale was carried out and the derived results are presented and discussed.