An Enhanced Toolset for JunoCam Images of Io with Interactive Mosaic Visualization and Segmentation-Enhanced Georeferencing

Introduction

Processing JunoCam's raw images of Io to obtain georeferenced images is a critical task in planetary science. Although JunoCam is not a dedicated scientific instrument, it has provided extremely valuable data about Jupiter and its moons, including Io. Producing accurate georeferenced images is especially important now, as the extended mission aims to further study Io and the other Galilean moons.

 

Io, the innermost of Jupiter's four Galilean moons, is the most geologically active body in the Solar System, with over 400 active volcanoes [1]. This volcanic activity is driven by tidal heating resulting from Io’s orbital resonance with Europa and Ganymede. Observations from JunoCam provide crucial insights into Io's surface composition, volcanic activity, and interior structure.

 

JunoCam

JunoCam is a CCD sensor equipped with four filters red, green, blue, and methane designed to capture detailed color images of Jupiter and its moons [2]. JunoCam raw images are publicly available in PNG format and composed of multiple framelets of 128 x 1648 pixels, for each RGB filter. The reconstruction of these framelets, the optical distortion corrections, and the segmentation of the target of interests, are fundamental steps to achieve precise georeferenced results. The limb identification process is pivotal for creating accurate map-projected images. Identifying the limb accurately in each framelet allows for precise reconstruction and mapping of the image onto a georeferenced coordinate system.

 

Workflow Description

In the proposed workflow, raw images are enhanced through multiple Computer Vision (CV) pre-processing steps, to improve contrast and reduce noise, making limb detection more robust. Edge-detection is first applied to the enhanced images, to filter-in the framelets containing the target of interest. Then, segmentation masks are obtained by processing the framelets with the Segment Anything Model (SAM) [3]. Since SAM is class-agnostic, all the obtained masks are filtered by combining the overlapping ones and checking the corresponding DN mean values. This step is crucial for undistorting and map-projecting the framelets accurately.

The segmented framelets are then processed using a pin-hole camera model and SPICE kernels to undistort and map-project them. SPICE kernels provide spacecraft and planetary ephemeris data, which are essential for accurate map projection. The undistorted framelets are then stitched together to reconstruct the final georeferenced image.

 

Results

The proposed workflow has been tested on several JunoCam images of Io. The results show a good accuracy of georeferenced images, with a reproducible and robust workflow. The limb segmentation step, powered by SAM, is particularly effective in isolating the limb despite the challenging conditions present in raw space images, such as varying lighting and noise levels.

 

Figure 1. Interface of the interactive plot where users can create and interact with reconstructed uncontrolled mosaics of Junocam images. This tool allows for real-time adjustments and visualization, enhancing the user's ability to fine-tune the reconstruction process.

Figure 2.  Framelet segmentation step: the original raw image (a) and the raw image with segmentation masks superimposed (b). The segmentation masks, shown in red, highlight the detected limb, which is crucial for accurate georeferencing.

Conclusion

The semi-automated robust workflow presented in this paper significantly improves the georeferencing of Junocam images of Io. By leveraging advanced segmentation techniques and adhering to the FAIR principles, this work provides a valuable tool for the planetary science community. Future enhancements could include further automation and integration with other planetary image processing tools.

 

FAIR Principles

This work adheres to the FAIR principles—Findability, Accessibility, Interoperability, and Reusability—by publishing the workflow code as open-source, ensuring it is easy to use and modify. 

Acknowledgments

This work is supported by the Agenzia Spaziale Italiana (ASI). JIRAM is funded by the ASI–INAF Addendum n. 2016-23-H.3-2023 to grant 2016-23-H.0. Part of this work was performed at the Jet Propulsion Laboratory, California Institute of Technology, under contract with NASA.

 

References

[1] Zambon, F., Mura, A., Lopes, R. M. C., Rathbun, J., Tosi, F., Sordini, R., et al. (2023). Io hot spot distribution detected by Juno/JIRAM. Geophysical Research Letters, 50, e2022GL100597. https://doi.org/10.1029/2022GL100597

[2] Hansen, C.J., Caplinger, M.A., Ingersoll, A. et al. Junocam: Juno’s Outreach Camera. Space Sci Rev 213, 475–506 (2017). https://doi.org/10.1007/s11214-014-0079-x

[3] Kirillov, A., Mintun, E., Ravi, N., Mao, H., Rolland, C., Gustafson, L., ... & Girshick, R. (2023). Segment anything. In Proceedings of the IEEE/CVF International Conference on Computer Vision (pp. 4015-4026).