A Data Fusion Technique for Mosaicking of Different Sources Digital Elevation Models

Digital elevation models (DEMs) can be obtained by using different techniques, either based on measurements in situ or remote sensed data (e.g. levelling, photogrammetry, SAR interferometry, radargrammetry, laser scanning, etc.). Different DEMs are usually not homogeneous and affected by different systematic vertical and horizontal errors, in addition to random noise. Standard mosaicking procedures try to reduce only the inconsistencies in overlapping areas and provide results where discontinuities are no more clearly visible, but do not remove the systematic errors that caused the artefacts. In this work we propose a method that exploits the information contained in the area of overlap between different DEMs in order to reduce horizontal and vertical systematic errors. Then, after systematic errors have been removed, more standard mosaicking methods are used to reduce random noise and fill areas where data are missing. The same fusion approach is suitable for mosaicking other kinds of images in addition to DEMs. The proposed technique allows obtaining accurate and homogeneous DEMs, as demonstrated in the framework of the DUDES project funded by ESA. The DEM obtained from the fusion of ERS and SRTM interferometric DEMs has been validated by comparison with a high resolution DEM, and largely fulfils DTED2 specifications.