Lack of point cloud data in well aligned areas.

[AllanMartin]

Hi All,
I'm new to the forum as my company has only just branched into using Agisoft.
Basically  we're looking an generating and orthophoto of a quarry floor. And to do this we've used a calibrated panoramic camera head taking sixteen images at a time. We've grouped each set of photo's into stations, and have generated a point cloud.
We tested this in a car park and had excellent results (Test 1 attached) but when we tested in the quarry, we've been unable to get the images to align, after 40 tests the best we have are 19 of out 25 stations aligning, and we had to put 24 markers in to get that. But the main problem is a lack of point cloud on test 2.
As you can see from the attachments we've gone from 800 cameras to 400, but we've gone from 34.5m tie points to 1m.
We are using alignment settings of High ; Reference ; 40000 ; 1000-10000 ; Adaptive fitting. And we have put in the camera locations and yaw pitch and roll of each camera.
Any help would be greatly appreciated!
Thanks

[Alexey Pasumansky]

Hello AllanMartin,

I can suggest to check the overlap for the images from the neighboring camera stations (for problematic and aligned stations), you can try to run the alignment only for three stations (two aligned and one - not aligned) that that closest to each other.

For processing workflow I can recommend to uncheck all the cameras in the Reference pane and re-run the alignment using Generic preselection.

Also if you are interested in the model and orthomosaic of the floor only, then it may be sufficient to take only vertical images with the sufficient overlap, similar to the aerial surveys, unless the technique you've used for the project is the one you are already used to and have multiple datasets captured in a similar way.

[AllanMartin]

Thanks for your reply Alexey!

Unfortunately we're tied to the workflow, but I will definitely test your processing settings.

We've tested the overlap extensively and that should be okay, but I'll try running three stations together.

One thing we've noticed is the Quarry images are much brighter than the car park. Could this be something thats causing us problems?

Many thanks,
Allan

[Photogrammetryfacts]

This is just my anecdotal evidence, but I've found that underexposing overexsposed images by a few stops and increasing the contrast can do wonders for data sets that otherwiser have problems aligning. So that could at least be worth a try perhaps?