Update rs2dataset.cpp #1668
Update rs2dataset.cpp #1668
Conversation
The Affine GeoTransform is wrongly calculated , while the "positioningInformation" being assumed as AREA_OR_POINT=POINT (point coordinates from a "mapCoordinate" item is assumed to be a center of the pixel). But really the map projection coordinates from "positioningInformation" are AREA_OR_POINT=AREA (the coordinates of pixel corners). That could be proved by equal coordinate values of these point from "geodeticCoordinate" data and corresponding coordinates of GCP data. As a result of that, georeferencing performed from GeoTransform and GCP point data are different with a half pixel shift and slightly wrong scaled.
|
Are you really sure about that ? |
|
Yes, you are right. According to the document https://spacedata.copernicus.eu/documents/12833/14537/Radarsat-2_ProductFormatDefinition Chapter A "IMAGE CO-ORDINATE REFERENCE SYSTEMS" all coordinates of the product are related to PixelIsCenter, except direct GeoTiff goereferencing which is corrected to be PixelIsArea. |
You can just update the issue and this pull request (git commit --amend and force push) |
|
There are doubts appeared, do the dfGCPPixel and dfGCPLine coordinates belong to the AREA_OR_POINT=Area model. I can not find the information in GDAL manuals. Or it should follow a model defined in GDALMD_AREA_OR_POINT, being AREA_OR_POINT=Area as default? |
I guess so. This might not be explicited in the doc, but this is the practice in other drivers and how the GCP transformer works as far as I can remember. AREA_OR_POINT=Area is the dominant model in GDAL. AREA_OR_POINT is mostly a metadata hint in GDAL to indicate format encoding convention (for example in the GeoTIFF driver, this will influence whether a half-pixel shift is done between GDAL GeoTransform and GeoTIFF tiepoint) that has no influence on GDAL algorithms. |
What does this PR do?
Radarsat 2 driver: correct affine GeoTransform calculation (fix #1666)
What are related issues/pull requests?
#1666