ScenProc manual section 6 says...
Sometimes you want to create autogen for an area that is not covered by good
vector data. So for example the vector data shows hardly no vegetation, while
in your photo scenery you can obviously see the vegetation. How to tackle this
problem?
The feature detection function of scenProc is designed for that. With this
feature you can create your own vector data, derived from the imagery you use
for your photo scenery. This vector data can then be used later on to create
your autogen like usual.
Does anyone have experience of this?