havardlovas
commented
1 year ago Created branch feature_parser to handle this issue. Will be merged with main once it is working.
Closed havardlovas closed 9 months ago
havardlovas
commented
1 year ago Created branch feature_parser to handle this issue. Will be merged with main once it is working.
The current parsing is poorly structured, and ideally the parsing layer should deliver the following in a standardized form:
Vehicle position/orientation (pose): The preferred format is pose with respect to earth-centered earth-fixed (ECEF) coordinate reference system. This with either come from visual navigation where the camera defines the body frame, or from an GNSS-INS where the IMU typically defines the body frame.
HSI geometry: Since we work with push-broom sensors, the geometry includes the line camera model parameters and the orientation/position of the sensor. These are also called boresight/lever arms.
3D triangle mesh (mesh.ply): The 3D mesh is used for ray intersection a.k.a. ray casting/tracing. It may be derived from a DEM or a mesh from photogrammetry.
Data cube: Raw data cubes may come in hdf formats or ENVI formats. Processing ENVI cubes into hdf-format with embedded NAV seems useful. This also allows the handling of oversized data cubes.
Reference Orthomosaic/DEM (optional): For validation/calibration it is important to have reference orthomosaics/DEMs. This data can be downloaded from online databases (other surveys) or from a photogrammetry project. Important to be aware of the vertical reference of the DEMs. It is on the TODO to to add support for both ellipsoid-relative and geoid-relative elevations. Be sure that the DEM and vehicle positions are at least consistent (e.g. if derived from water depth + tide correction).
[ ] Create a module parsing_tools containing the relevant tools to give a standardized format of data to be used for georeferencing. Parsing could be a script that uses the config file to handle the variety of "raw data formats".
[ ] 1. Parse (if necessary) data cube formats into .h5 with size-control
[ ] 2. Parse various pose formats to pose-per-HSI frame and embed as folder in .h5 file. Must be done in conjunction with the above.
[ ] 3. (Independent) Parse (if necessary) HSI camera geometry into *.xml file. Maybe do something with the annoying binning stuff for UHI.
[ ] 4. (Independent) Parse 3D model into a mesh. It is important that potential offsets are applied. In future, simplistic altimeter registration (underwater) or constant (altitude) modes should be implemented.