Aerial imagery 'holes' in multiple data sets

[readytowalk]

Hello all, thanks for taking the time to read this.

I have 2 imagery data sets (CIR and RGB) of the same field (~125 acres).  When I processed them through PS, they both showed 'holes' in nearly the same place where cameras did not align and tie to surrounding images.  The CIR set had 12 cameras that did not align and the RGB had 47.  The rest of the mosaic looks terrific and exported at 2cm (the files attached are a screen grab at 25cm) so I do not think the aircraft log file was erroneous (GPS loss, etc) or anything like that.  If it helps, the camera is a NEX7 at 12mp and set for 65/65 overlap and sidelap.  Also, I am not using GCPs.

I am scratching my head with this one and can't come up with any theories. Any direction or suggestion would be helpful. 

[mobilexcopter]

Hello,

I think it happens when Photoscan can't tie the points together, mainly on vegetation, trees,... It happens to me also. It might help if you would use more than 65% overlap. I think around 80% overlap and 60%+ side overlap is recommended.

Best regards,

Alex

[HyperFox]

Hello,

obviously, the wrong image matching results from a poor overlap. I recommend strongly to use at least 70 to 75 % . However, above Vegetaion even 80% will not be enough, because of object movements and a low contrast / dark regions with iso noise etc. If you have more matches in NIR than in VIS, this is a strong hint to problems with vegetation since vegetation shows much more contrast in NIR imagery.

I'm just testing 90% Overlap, will post here the next days.

regards
Ansgar

[David Cockey]

What accuracy settings are you using for Align and Build Dense Cloud? I have found with a few photo sets to be very sensitive to those settings. Try reprocessing the photos near the holes as a separate chunk with different accuracy settings.