Increasing accuracy when using Full Motion Video

Without jumping right into the nitty gritty of how full motion video works—I’ll do that below for the geeks who read this blog—you can use an elevation dataset to increase positional accuracy when working with FMV. To access the Accuracy dialog go to the Video Manager >> View >> Accuracy. Here is an elevation dataset from ArcGIS Online. Here’s where you can get a run down of FMV and download it.

Key FMV concepts to keep in mind:

You have videos that are “spatially aware.” This doesn’t mean that every frame in the video has geospatial information which is used to get everything to line up correctly. Every x number of frames will have information on the four corners and frame center, and then the spatial information for the other frames is estimated based on the movement of the aircraft, sensor, etc. It’s pretty good on its own at being in the right place; but, with all of that estimation occurring, it’s possible to propagate error.

The other thing to keep in mind is that the earth is neither flat nor round. It’s not even spherical. It’s a big rock, which happens to look kind of like an ellipsoid. An ellipsoid with a bunch of holes and things sticking out of it. And you’re trying to get a video shot from a wobbly airplane to line up right…just keep that in mind. Even a slight change in the viewing angle can have a tremendous effect on where features appear on the map.

In short, because of the limits of the metadata and the surface onto which the videos are projected, you can reduce the amount of location error by incorporating an elevation dataset.

The only caveat to this is that it requires more computational power, so you may have to decide if the trade off in increased accuracy is worth the performance. There are a lot of those kinds of decisions you have to make in FMV because it asks a lot of your machine. Anyway, here’s the promised nitty gritty:

Polynomial Transformations versus Camera Models

Typically a polynomial transform is created for each video image based on the coordinates of the four image corners. This transformation is inherently approximate, but can be useful in the absence of other information.

Another approach is to create a camera model based on the video metadata and elevation data.  The camera model considers the aircraft position and orientation, camera look angles, lens settings and ground elevations.  If all this information is available and accurate the camera model can potentially be more accurate than the polynomial method.

When you select an elevation source in the Accuracy dialog, you are directing the program to use the camera model calculations when digitizing points from a video frame.  The elevation source may be a DEM (digital elevation model) or a constant elevation.

The camera model uses the ECEF (Earth Centered Earth Fixed) coordinate system in intermediate steps when calculating the spatial relationships between the aircraft and ground locations.  This coordinate system expresses locations near the earth’s surface in a 3D Cartesian space, simplifying distance and direction calculations.  All of this is done behind the scenes, and the resulting geographic coordinates can be projected into any map projection as usual.

It is important to note that the accuracy of the camera model is highly dependent on the presence and accuracy of the metadata and terrain.  Consequently you may not see improvement when specifying an elevation source and using the camera model.  In these cases the camera model option should not be used.

This entry was posted in Analysis & Geoprocessing, ArcGIS Online, Defense, Imagery, National Government, Petroleum, State Government and tagged , , , . Bookmark the permalink.

Leave a Reply