Day 4

Detailed paper information

Back to list

Paper title LibGEO, Space imagery Geometry Library for Operational Missions and Expert Studies
  1. Céline L'Helguen Centre National d’Etudes Spatiales (CNES), Toulouse, France
  2. Nicolas Champdavoine Magellium Speaker
  3. Fabrice Buffe Centre National d’Etudes Spatiales (CNES), Toulouse, France
Form of presentation Poster
  • B6. National missions TPM
    • B6.01 National EO satellite missions
Abstract text LibGEO is a multi-sensor geometric modeling library, with high location precision. The library is designed to be used at different steps of an Earth Observation mission: prototyping, ground segments, calibration. It is first developed to meet CNES/Airbus Defense and Space, CO3D mission requirements and then to be the CNES reference library for geometry.

The base function of LibGEO is the direct location which returns ground coordinates for each pixel coordinate of the image. It supports both mathematical (grid or RPC) and physical modeling. For physical modeling, a line of sight is built from the detector and is transformed using, for example, rotation, translation, homothety, mirror reflection, to get the line of sight in the International Terrestrial Reference Frame (ITRF). With the position of the platform and an ellipsoid model of the earth, the ground position can be computed. Other location functions are provided such as inverse location, intersection on DEM, colocation. Each location function has a grid implementation, which creates grids to resample images in different geometries (including orthoimages). LibGEO also deals with stereo images for 3D model reconstruction. It has the ability to intersect lines of sight from correlated image points to get a 3D point. LibGEO computes the epipolar geometry that allows dense correlation of the 3D reconstruction.

Native location is not precise enough for some applications, therefore an optimization of the model parameters can be done by using Ground Control Point (GCP) and tie points in image geometry. The improvement of absolute and relative location is done where the user can set uncertainties on both observations and model parameters. During the optimization process, points with higher residual errors are filtered out using statistical methods.

Supported sensors are Pleiades HR, Sentinel-2 MSI (L1B), CO3D for now and some will be added soon: Thrisna, 3MI/Metop SG, Microcarb. The library is built to be generic, other sensors could easily be supported by simply plugging in a product format handler.

The library is designed to be easily integrated in any operational processing chain thanks to its C++ API but it is also user friendly for prototyping and expertise through the Python API.