This is a repost from here with added detail.
Using a drone equipped with a video camera the goal is to create Full Motion Video. I wrote some software to convert the my drone's telemetry into a Video Multiplexor compatible format. So far so good.
It's ideal to use ESRI's Terrain world elevation layer for quality and convenience, but there's a lot of uncertainty in the documentation and tool output.
For example, if you use the Terrain elevation layer as input for analysis with no definition query, there is no way to determine what underlying elevation dataset was used as tool input. The docs page tells us that the default behavior optimizes for "Best", where "Best" means smallest cell size (highest resolution).
This is inherently problematic if you care about data integrity. Were NAVD values used? EGM96? An uninformed user might use a tool like Extract Values to Points and have no idea that the vertical datum could be something unexpected depending on what "Best" means for their geographic area.
If you use the "default" Terrain service as DEM input to the Video Multiplexor tool, then you automatically encounter the problem detailed above. But you also encounter several compounding problems depending on flight altitude, relative camera angle, and camera internal characteristics (such as HFOV and VFOV). The Video Multiplexor requires a DEM that covers the entire flight area and calculated FOV footprint. This means that if you have a video taken at nadir, but even 1 frame shoots above the horizon, the tool still requires an almost continental DEM to succeed. Adding to this issue, the error message doesn't tell you the actual nature of the problem. It just tells you that it failed to extract elevation values from the DEM. Definition queries for semi-global DEMs such as STRM do not guarantee success here, because if the horizon appears in the video then the footprint is theoretically infinite. The docs do not explain how that extent is calculated.
I did lots of tests, such as forcing every frame's relative camera angle to each integer from 0 to -90. I could find no common denominator for success.
Finally on a whim, I tested the Terrain service's World_0_M dataset as definition query input to the Video Multiplexor tool. This input works in every case, but I can't find any documentation or metadata for this layer. The name implies a global fictional zero elevation raster. If you click to identify raster values in the Map it always returns 0 meters elevation. But astonishingly, the output FOV footprint shows elevation contours. What is going on here? I can guarantee Video Multiplexor success, but I have no idea what the underlying vertical datum or dataset is. Any links or clarity are much appreciated.