Extract Video Frames to Images tool mixes output coordinate systems in Corner Coordinates

[Anonymous User]

Extract Video Frames to Images tool mixes output coordinate systems in Corner Coordinates.

After running the tool, I inspected some of the output files with gdalinfo.

The Corner Coordinates data appears to be a mix of lat/lon and projected coordinates. Is this the intended behavior?

[CodyBenkelman]

I've asked our engineer responsible for this tool to comment - so we may update this response - but I looked at an extracted video frame (tif format) using GDALINFO and it appears correct (below) to me - I highlighted the corner coordinates in bold - note these are raw image pixel coordinates, not a projected coordinate system.  

c:\image_mgmt_workflows\OptimizeRasters\GDAL\bin>gdalinfo C:\Users\cody6520\Documents\ArcGIS\Projects\ProProject_FMV+OIC_UC2022droneSeminar\ExtractFrames\YUN_0127_multiplexed_GimbalEdited_0000004.tif

Driver: GTiff/GeoTIFF

Files: C:\Users\cody6520\Documents\ArcGIS\Projects\ProProject_FMV+OIC_UC2022droneSeminar\ExtractFrames\YUN_0127_multiplexed_GimbalEdited_0000004.tif

Size is 1920, 1080

Coordinate System is:

COMPD_CS["unknown",

    GEOGCS["GCS_WGS_1984",

        DATUM["WGS_1984",

            SPHEROID["WGS_84",6378137.0,298.257223563]],

        PRIMEM["Greenwich",0.0],

        UNIT["Degree",0.0174532925199433],

        AUTHORITY["EPSG","4326"]],

    VERTCS["WGS_1984",

        DATUM["D_WGS_1984",

            SPHEROID["WGS_1984",6378137.0,298.257223563]],

        PARAMETER["Vertical_Shift",0.0],

        PARAMETER["Direction",1.0],

        UNIT["Meter",1.0],

        AUTHORITY["Esri","115700"]]]

Origin = (0.000000000000000,0.000000000000000)

Pixel Size = (1.000000000000000,-1.000000000000000)

Image Structure Metadata:

  INTERLEAVE=PIXEL

Corner Coordinates:

Upper Left  (   0.0000000,   0.0000000)

Lower Left  (       0.000,   -1080.000)

Upper Right (    1920.000,       0.000)

Lower Right (    1920.000,   -1080.000)

Center      (     960.000,    -540.000)

Band 1 Block=1920x1 Type=Byte, ColorInterp=Red

  Overviews: 960x540, 480x270, 240x135

  Metadata:

    COLORINTERP=Undefined

Band 2 Block=1920x1 Type=Byte, ColorInterp=Green

  Overviews: 960x540, 480x270, 240x135

  Metadata:

    COLORINTERP=Undefined

Band 3 Block=1920x1 Type=Byte, ColorInterp=Blue

  Overviews: 960x540, 480x270, 240x135

  Metadata:

    COLORINTERP=Undefined

Are you asking because you're having some problem with the images?  A few important points:

1) if your video is aimed above the horizon, exporting a frame may result in a cropped image

2) these video frames are not orthorectified - ArcGIS is not attempting to apply a lens model, nor fit the image to terrain - so it's not intended to support measurements, but simply provide a convenient visualization.  

Let us know if that answers what you need

Cody B

[Anonymous User]

Thank you for the detailed response on this. When I multiplex the videos and play them in Pro, the footprint geometry looks pretty good. But the corner coordinates from the tool export are interesting to say the least (the videos weren't take in the middle of the ocean).

Here's a sample:

Size is 1920, 1080
Coordinate System is:
GEOGCRS["WGS 84",
    DATUM["World Geodetic System 1984",
        ELLIPSOID["WGS 84",6378137,298.257223563,
            LENGTHUNIT["metre",1]]],
    PRIMEM["Greenwich",0,
        ANGLEUNIT["Degree",0.0174532925199433]],
    CS[ellipsoidal,2],
        AXIS["geodetic latitude (Lat)",north,
            ORDER[1],
            ANGLEUNIT["degree",0.0174532925199433]],
        AXIS["geodetic longitude (Lon)",east,
            ORDER[2],
            ANGLEUNIT["degree",0.0174532925199433]],
    ID["EPSG",4326]]
Data axis to CRS axis mapping: 2,1
GeoTransform =
  58.59252494274197, 2.162131309546567e-07, -117.1850503877213
  -16.59005182917486, -1.071558551224427e-06, 33.18010424409931
Image Structure Metadata:
  INTERLEAVE=PIXEL
Corner Coordinates:
Upper Left  (  58.5925249, -16.5900518) ( 58d35'33.09"E, 16d35'24.19"S)
Lower Left  ( -126501.262,   35817.923) (Invalid angle,Invalid angle)
Upper Right (  58.5929401, -16.5921092) ( 58d35'34.58"E, 16d35'31.59"S)
Lower Right ( -126501.261,   35817.920) (Invalid angle,Invalid angle)
Center      (  -63221.334,   17900.665) (Invalid angle,Invalid angle)
Band 1 Block=1920x1 Type=Byte, ColorInterp=Undefined
  Overviews: 960x540, 480x270, 240x135
Band 2 Block=1920x1 Type=Byte, ColorInterp=Undefined
  Overviews: 960x540, 480x270, 240x135
Band 3 Block=1920x1 Type=Byte, ColorInterp=Undefined
  Overviews: 960x540, 480x270, 240x135

Not sure what's going on here but the comments are greatly appreciated. This isn't a critical issue. I was just trying to squeeze more data out of the exports since the csv output doesn't include all MISB fields.

[NickWallingford · Accepted solution]

Ooof. Yup, that's a bug.

We have a fix for 3.1. We've missed the window for 3.0.1, but I'll try to get the fix in 3.0.2.

Unfortunately there's no workaround.

[Anonymous User]

Thanks for the follow up on this - it's much appreciated.

[NickWallingford]

As a further follow up, the bug has been fixed for 3.0.2, currently scheduled to be released late September.

Thanks again for reporting this issue.

[Anonymous User]

For posterity, I tested this tool with the ESRI FMV tutorial data and still have mixed and incorrect Corner Coordinate output.

No solid evidence, but I think the issue here might be the DEM. I'm using the ESRI Terrain service for DEM input.