flight path affect on reprojection error

[bisenberger]

I flew project with north-south flight lines and then flew it again with east-west flight lines. If I combine all the photos into one chunk the reprojection error is around 1pix. If I make two chunks one for the north-south photos and one for the east-west photos the reprojection error is around .7 pix for both chunks. The higher reprojection error has an adverse affect on control point error.

[stihl]

Have you tried filtering the tie-points with gradual selection (reprojection error, projection accuracy, reconstruction uncertainty and image count) so that the reprojection error is around .7 pixels?

[ekbmuts]

stihl,  I use the gradual selection tools frequently but don't really have a clue what they are doing exactly, nor what the best general settings may be.  I know that they eliminate erroneous points but not how.

My use of them is based on a writeup from a dinosaur guy, meaning, a paleontologist.  It's quite good actually.  But is there an official Agisoft brief on the use of these tools?

Jon

[bisenberger]

Have you tried filtering the tie-points with gradual selection (reprojection error, projection accuracy, reconstruction uncertainty and image count) so that the reprojection error is around .7 pixels?

To get it to around .7 pixels I have to remove about half of the original points 67,369 to 34,247. This gets my control point error to 0.076 meters (before filtering it was 0.165).

If I process them separately by flight direction, then merge the two chunks, the control point error is .0315 without any tie-point filtering. Total number of tie-points when merged is 64,502 (before merging one has 30,058 the other 34,444).

Another curious thing is I flew the same paths with different camera settings.  The paths flown with F-stop F/2, ISO 80 and shutter 1/1250 have the higher reprojection and control point error (listed above). The other camera settings I used were F-stop F/4, ISO 100, and Shutter 1/640. The reprojection error with these camera settings is 0.882 and the control point error is 0.033.

Here's a screen grab of the shaded mesh for the project:

[bisenberger]

stihl,  I use the gradual selection tools frequently but don't really have a clue what they are doing exactly, nor what the best general settings may be.  I know that they eliminate erroneous points but not how.

My use of them is based on a writeup from a dinosaur guy, meaning, a paleontologist.  It's quite good actually.  But is there an official Agisoft brief on the use of these tools?

Jon

Hi Jon,
Check this thread:
http://www.agisoft.com/forum/index.php?topic=1359.0