CARS, a satellite multi view stereo framework

Overview • Quick Start • Documentation • Contribution • References

Overview

From stereo images	CARS produces a Digital Surface Model (DSM)

CARS is an open source 3D tool dedicated to produce Digital Surface Models from satellite imaging by photogrammetry. This Multiview Stereo framework is intended for massive DSM production with a robust, performant and modular design.

Be aware that the project is new and is evolving to maturity with CNES usage roadmaps and projects such as:

• CO3D project    
• AI4GEO project    
• Evoland project    

Quick start

CARS Docker Image

CARS is available on Docker Hub and can be downloaded by:

docker pull cnes/cars

One main pipeline to generate DSM

You only need to launch one command:

docker run -w /data -v "$(pwd)"/data_gizeh:/data cnes/cars /data/configfile.json

with one configuration input file ("configfile.json") located in a "data" folder to be consistent with the previous command lines:

{
    "inputs": {
        "sensors" : {
            "one": {
                "image": "img1.tif",
                "geomodel": "img1.geom"
            },
            "two": {
                "image": "img2.tif",
                "geomodel": "img2.geom"
            }
        }
    },
    "output": {
        "directory": "outresults"
    }
}

On the way to the Pyramids...

You want to build the pyramids by yourself? Download our open licence Pleiades data sample to give CARS a try!

Documentation

Go to CARS Main Documentation.

Contribution

To do a bug report or a contribution, see the Contribution Guide.

For project evolution, see Changelog.

Credits

See Authors file

References

• Youssefi D., Michel, J., Sarrazin, E., Buffe, F., Cournet, M., Delvit, J., L’Helguen, C., Melet, O., Emilien, A., Bosman, J., 2020. Cars: A photogrammetry pipeline using dask graphs to construct a global 3d model. IGARSS - IEEE International Geoscience and Remote Sensing Symposium.
• Michel, J., Sarrazin, E., Youssefi, D., Cournet, M., Buffe, F., Delvit, J., Emilien, A., Bosman, J., Melet, O., L’Helguen, C., 2020. A new satellite imagery stereo pipeline designed for scalability, robustness and performance. ISPRS - International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences.