Open-source 3D tools developed for the CO3D mission and beyond

After several years of development, mid-2025 should see the launch of the CO3D¹ (Optical Constellation in 3D) mission. The CO3D project is a public-private partnership between CNES (French space agency) and AIRBUS (aerospace company).

The mission aims to launch pairs of optical satellites to photogrammetrically reconstruct the Digital Surface Model (DSM) of the Earth's land surface. A notable innovation is the ability to manage stereo acquisition in synchronous mode for capturing moving elements.

As part of the CO3D program, CNES is developing the 2D and 3D product processing chains integrated into the mission's ground segment, alongside an Image Calibration Center (ICC) for radiometric and geometric calibration. These systems leverage a suite of open-source tools created by CNES teams. CNES chooses to release these tools prior to the mission's launch to gather user feedback and improve the quality of final CO3D products.

• CARS² is the Multiview Stereo Framework (MVS). It produces DSM from satellite images acquired from different view angles.
• PANDORA³ is the stereo matching framework. It combines both switchable similarity measures at the pixel level and regularizers. The CARS pipeline integrates PANDORA.
• BULLDOZER⁴ removes above-ground elements (e.g., trees, buildings) from DSM generated by CARS to produce Digital Terrain Models (DTM).
• XDEM⁵ evaluates and validates the 3D quality of the generated models (i.e., the CARS DSM and the Bulldozer DTM).
• SLURP⁶ produces land cover masks from very high-resolution (VHR) images. The CARS pipeline integrates these masks to generate sharper 3D reconstruction.

CO3D mission represents a significant step forward in Earth observation, offering innovative tools for producing accurate digital surface models. By engaging with the user community early, CNES aims to ensure the mission delivers results that fit user needs.

[1] Lebegue, L. et al. (2024). CO3D Products Qualification Forecast. IGARSS 2024 - 2024 IEEE International Geoscience and Remote Sensing Symposium, 8555-8559.
[2] Youssefi, D. et al. (2020). CARS: A Photogrammetry Pipeline Using Dask Graphs to Construct A Global 3D Model. IGARSS 2020 - 2020 IEEE International Geoscience and Remote Sensing Symposium, 453-456.
[3] Cournet, M. et al. (2020). Ground-truth generation and disparity estimation for optical satellite imagery. ISPRS - International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences.
[4] Lallement, D. et al. (2023). Bulldozer, a free open source scalable software for DTM extraction. The International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences.
[5] Hugonnet, R. et al. (2022). Uncertainty Analysis of Digital Elevation Models by Spatial Inference From Stable Terrain. IEEE Journal Selected Topics in Applied Earth Observations and Remote Sensing, 15, 6456–6472.
[6] Tanguy, Y. et al. (2024). Smart Land Use Masks: A Simple and Robust Approach to Produce Low/High Vegetation Masks from a Single High Resolution Satellite Image. IGARSS 2024 - 2024 IEEE International Geoscience and Remote Sensing Symposium, 4164-4168.