01-23-2017 03:43 AM
I have a .blk with two historical frame camera images. I have already done the triangulation and I have tie points and GCP.
I would like to use SGM with this images. I know that it is not possible. I'm looking for a workaround. Can I create a new blk with digital camera as sensor and then load the frame camera images and the points? Will it work?
Solved! Go to Solution.
01-23-2017 05:38 PM
Have you considered using classic ATE or eATE terrain extraction tools instead of SGM? Both will generate a surface using frame sensor geometric model. eATE can also generate a dense point cloud. You won't need to create a new digital camera block file if you use eATE.
As per your question it may work. You need well distributed highly accurate GCPs. If your triangulation results are good with frame camera block file, you may turn the tie points into GCPs and import them in digital camera block that you will generate from scratch.
For frame cameras interior orientation of the camera is defined by fiducial marks, and you will lose this information. You need to define a new camera file, you may use focal length and principal point coordinates that are given in the camera calibration report. You need to compute the pixel size. First check your camera calibration report, in some calibration reports principal point coordinates are given both in pixels and millimeters. Using these values you may compute the relation between pixels and mm. Actually every other value given in pixels and mm should have the same scale difference. You can also compute the pixel size from solved interior orientation of the frame camera using film and image coordinates.
01-24-2017 02:25 AM - edited 01-24-2017 02:26 AM
Just to give you added confidence, one of my customers has successfully used XPro SGM with analogue data essentially using the process Yasemin described. He had to make a best guess as to pixel size and then performed a space resection to account for that unknown. Good ground control is a must. He got really good dense point matching. Cheers, Johnnie
02-09-2017 04:38 PM
I computed the pixel size in x=66.86 microns and in y=65.95 microns. I computed the distance between fiducial 2-1 in pixel and in image coordinate systems then I computed the relation between pixel and mm to find the pixel size in x. Similarly I used the distance in y between fiducials 3-4 to compute the pixel size in y.
Alternatively, you may open the block file with interior orientation calculated for the frame in HFA Viewer as shown in the screen capture below, the numbers are already there (via HFAView, 2.5044427823076296e-002 mm for x, 2.5053843378096682e-002 mm for y).