I would like to understand the geometric analysis a bit better. My current understanding is that it is the same as the normal control point generation only that it is constrained to searching only in overlapping images for control points. Is this correct? I have been struggling w/ a monster pano (6000 images) I was hoping that the geometric analysis would help but it ran for 5 days and stopped progressing, meanwhile the normal control point generation ran overnight. Can you explain a little about the difference between these two methods so I can better understand when to apply it?
Sure. In fact, you are right, it should behave the same and doesn't as in the standard detection mode, it is fully multithreaded and in the geometrical analysis, it is not at all. This is one difference. The second difference, is that during geometrical analysis, the settings of detection are not the same as in default detection, as we use the knowledge of overlapping to inspect better than with initial detection. I seldom use the global geometrical analysis on big panorama, but quite often do local geometrical analysis to fix only the location on the gigapixels where there are issues ( select some images, right click, local geometric analsysis ). This way, by looking around the panorama in one hour, you probably can fix everything. I did that method on paris project and it worked quite well ( I also used manual move as some images don't have CP as they were blurred )