metadata model in OTB v8

OTB v8 has a new metadata management which will lead to significant changes in diapotb.

A metadata model based on GDAL and not on OSSIM/OssimPlugins anymore, will be available in OTB v8.

Several impacts can be quoted :

• Ossim dependancy is removed, see : #7
• New management based on GDAL :
  • New keywords and APIs
  • Metadata included in tiff image (or with .aux.xml) instead of geom file

See : https://gitlab.orfeo-toolbox.org/orfeotoolbox/otb/-/wikis/Migration-guide-OTBv8

Impact on applications

Several applications used metadata to get or set information.

The main refactoring for metadata inside DiapOTB will be to get/set keys with right names and by using Set/GetImageMetadata instead of Set/GetImageKeywordlist.

Some examples available in https://gitlab.orfeo-toolbox.org/orfeotoolbox/otb/-/wikis/Migration-guide-OTBv8#migration-example

In OTB v8, keys are several kinds :

• MDNum for numeric keys (double)
• MDStr for string keys
• MDTimes for dates, the Time struct is defined here
• MDGeom for models (RPC model, Sar model, GCPs ...)
• MDL1D et MDL2D for LUTS, the LUT struct is defined

Quote all key names used in DiapOTB (from legacy imagekeywordlist):

• azimuthFmRate.*:
  • in SARParam.azimuthFmRates
• dopplerCentroid.* :
  • in SARParam.dopplerCentroids
• support_data.azimuth_steering_rate :
  • in SARParam.azimuthSteeringRate
• support_data.line_time_interval :
  • not always 1/PRF (ie : TSX) => perhaps keep both line_time_interval and PRF in SarParams.azimuthTimeInterval
• support_data.radar_frequency :
  • in v8 RadarFrequency
• support_data.geom.bursts.* :
  • SARParam.burstRecords
• support_data.range_sampling_rate :
  • SARParam.rangeSamplingRate
• support_data.slant_range_to_first_pixel:
  • SARParam.nearRangeTime
• support_data.first_line_time :
  • in v8 AcquisitionStartTime
• support_data.last_line_time:
  • in v8 AcquisitionStopTime
• support_data.abs_orbit :
  • in v8 OrbitNumber
• support_data.range_spacing :
  • in v8 PixelSpacing
• sensor:
  • in v8 Mission
• manifest_data.instrument:
  • in v8 Instrument
• support_data.orbit_pass:
  • in v8 OrbitDirection
• support_data.acquisition_mode:
  • in v8 Mode
• support_data.invalid_pixels (from SARBurstExtraction)
  • in v8 invalid_pixels
• support_data.number_samples:
  • in v8 NumberOfColumns
• number_lines :
  • in v8 NumberOfLines
• number_samples :
  • in v8 NumberOfColumns
• ul_lon/lat, ur_lon/lat ...:
  • SARParam.ulSceneCoord (or ur/lr/ll)
• support_data.azimuth_bandwidth:
  • SARParam.azimuthBandwidth

Tips: an easy way to match every metadata is to look at ParseGeom function in each *MetadataInterface. For instance : for Sentinel1

Impact on python_src

Python chains used some of metadata to check sensor, to apply some processing or update some metadata.

In v7, Several ways are used to retrieve/use metadata :

• Retrieve metadata as a dict from ReadImageInfo application with keywordlist = True
• Update metdata though a new geom file thanks to SARMetadataCorrection. A extended filename is used to take the new geom into account
• Read/Write geom file to get all metadata with expected time struct with python functions : read/write