Aerial photos with GPS, GCPs, limited RAM

[rossnixon]

I have looked at many posts, but it isn't obvious what approach to take.

I have about 4000 20MP photos (with GPS) to make into a high quality orthophoto (split into tiles).
I have 16GB of RAM, which I calculate is good for 600 photos at once.
I will use as many very accurate GCPs as required. Either 10 per chunk, or perhaps 20+ over the whole area.

How do I proceed? I see two choices.
Option 1: - process this as a series of overlapping chunks?
Option 2: - align all photos into one chunk using Reference preselection. I see mention of "duplicate the original chunk and generate mesh model for smaller areas by resizing the bounding box"

Any suggestions?

[rossnixon]

Option 2. was mentioned by Alexey Pasumansky here www.agisoft.com/forum/index.php?topic=2672.msg14160#msg14160

Does anyone know is this is the best of the two options when RAM is limited?

[Alexey Pasumansky]

Hello rossnixon,

I think 16 GB should be sufficient to align and optimize all the images at once. You can use 20 000 for key point limit and 4 000 for tie point limit to reduce the memory consumption mentioned here: http://www.agisoft.com/pdf/tips_and_tricks/PhotoScan_Memory_Requirements.pdf.

[rossnixon]

Thank you for that, Alexey.

In the following message, you mention a script for creating a grid of bounding boxes. Could I get this script please?
www.agisoft.com/forum/index.php?topic=778.msg16285#msg16285

[Alexey Pasumansky]

Hello rossnixon,

A sample for the version 1.2 is published on Wiki pages:
http://wiki.agisoft.com/wiki/Split_in_chunks.py