Updates to the TGO-CaSSIS Stereo Products Generation and the INAF public catalog of DTMs

In orbit since the April 2018, the Colour and Stereo Surface Imaging System (CaSSIS) on board of the ExoMars Trace Gas Orbiter (TGO) has recently entered its 6th year of scientific phase. Since then, CaSSIS has provided us with a rich catalog of up to four bands stereo images of the Martian surface. The stereo capability is achieved through an innovative telescope rotation approach composing a convergence angle of almost 22 degrees between the stereo couple. In this way CaSSIS can produce detailed color 3D maps, thus providing crucial data for the analysis of the surfaces and their composition.

The pipeline for three-dimensional modelling is developed and maintained by the INAF team located at the Astronomical Observatory of Padova, making available the Digital Terrain Models (DTMs) and orthorectified images to the entire team and scientific community since the start of the mission.

The 3DPD software (Simioni et al., 2021; Re et al., 2022), at the core of the pipeline, allows the exploitation of data according to the principles of stereogrammetry. The DTMs are produced at 13.5 m ground sample distance from 4.5 m/px images, with an estimated vertical accuracy below 3 pixels size (15 m)  (Fig.1).

To date, more than 2100 stereo couples are available on a total of 40.000 images acquired. Of these about 400 stereo pairs have already been processed and available for download at  the OAPD-hosted repository (https://cassis.oapd.inaf.it/archive/).

Since the framework was first founded by Simioni et al., 2021, the pipeline has been continuously developing to improve the performance of the data product generation.

The goal of this work is to present the actual state of the framework and all the improvements made. Recent changes are here described and supported by an assessment of the quality and precision of the generated products and their derivatives.

Recent developments include the integration of the Bundle Block Adjustment, employing the jigsaw routines made available with the USGS ISIS platform (Laura et al., 2023). Thanks to the jigsaw output, we are able to refine the projection matrices affected otherwise by distortions that introduce geometric effects of misalignments between the acquisitions. The misalignments not adequately modelled and resolved by the Bundle Adjustment could otherwise result in steps artefacts that can reach even hundreds of m in the worst cases.

Further important update is given by an innovative approach of aligning DTMs to MOLA-HRSC (Fergason et al., 2018), further improving the surface projection and the absolute elevation, reaching values generally below 50 m/px of standard deviation in comparison with it. This process was also extended to the entire database of DTMs already produced as one major update, bringing similar results (Fig.2).

Fig.1 Comparison between a CaSSIS DTM (MY34_003673_018) and a HiRISE DTM (DTEEC_005533_1975_005388_1975) at 1 m/px, demonstrating a vertical accuracy of about 8m.

 

Fig.2 The standard deviation on the vertical accuracy, achieved as a result of the recent alignment with the MOLA-HRSC and applied to the entire CaSSIS DTMs database.