Agisoft Metashape

Agisoft Metashape => General => Topic started by: aggieair on March 15, 2016, 10:41:45 PM

Title: Accuracies excpected for Historical Aerial Imagery DEMs?
Post by: aggieair on March 15, 2016, 10:41:45 PM
For those of you who have used scanned historical aerial imagery to create mosaics and surface models, can you share what kind of vertical accuracies you achieved and what were you data sources?

We have both black & white and color scanned images (at 1200 dpi) from 1948-2004.  I did not remove the fiducials/borders, so the image size extends to the end of the photo (this didn't pose too much of an issue actually).  I have no interior orientation info either.

For GCP markers, we pulled horizontal locations from a NAIP 1 meter resolution basemap, and vertical info from a LiDAR raster.  Finding points was definitely trick as this was Washington state where lots of clear cutting occurs so every year looks different, but we were able to find some control points.

Then we compared our DSM created from Agisoft to the existing LiDAR (ideally flat ground and avoiding the surface of the trees), and it's a pretty large difference.  Yes, we were kind of limited to the center of the images as the sides were super forested and steep inclines.  Depending on the scale (1:4800 and 1:1200), vertical ranged from 3 to 9 feet and horizontal ranged from 9 to 18 feet.

Just wanting to know if our results are as as good as expected or if getting RTK GPS would improve things.  I'm thinking not that much given the challenge of finding fixed features to go out and GPS that would be in our historical imagery and also the type of terrain so we're limited on how well to disperse the markers in the scenes.