Supporting Martian surface studies: a brand new dataset of 5cm Super-resolution restored ORI and 25cm Shape-from-Shading refined DTM

Publikation aus Digital

Y. Tao, J.-P. Muller, Laurence Tyler, Gerhard Paar

AOGS Asia Oceania Geosciences Society , 1/2016


Higher spatial resolution imaging data is always desirable to the  international community of planetary scientists interested improving understanding of surface formation processes. We have previously  developed a novel Super-resolution restoration (SRR) technique using Gotcha sub-pixel matching, orthorectification, segmentation, and 4th order PDE-TV, called GPT SRR [1]. This technique is able to restore 5cm-12.5cm near rover scale images (Navcam at range of ≥5m) from multiple 25cm resolution MRO HiRISE images [3]. The technique has been developed and successfully applied to MER and MSL missions at MSSL. In parallel, the Large Deformation Optimisation Shape From Shading (LDOSFS) algorithm, developed by Aberystwyth University (AU), applies calculated deformations on a succession of scales to the initial surface, gradually refining the level of detail in the resulting DTM [4]. Using an initial coarse DTM generated by stereo matching as a “seed” surface for the SFS algorithm, together with a higher-resolution image, it is possible to refine the original DTM and bring out fine surface detail. In PRoViDE, the SFS algorithm has been applied to an ORI at 25cm generated from an initial 1m resolution stereo-HiRISE DTM of three test sites on Mars, i.e. Homeplate for MER-A, Victoria Crater for MER-B and Shaler for MSL. By combining the 25cm SFS refined DTM with a 5cm SRR ORI, we are able to study the Martian surface with 5x-enhanced resolution on texture and 4xenhanced resolution on heights. These brand new datasets will be demonstrated in an interactive webGIS system (PRoGIS) and 3D Viewer (PRo3D®) both developed within the EU FP-7 PRoViDE project [2]. In the future, we plan to apply the same techniques for the ExoMars Trace Gas Orbiter 2016 CaSSiS instrument (from 4m up to ≤1m/pixel) including both 3D and SRR images from multiple overlapping colour stereos.