Bug: Photoscan crashes on Alignment of a large nr frames (4D tools)

[Meije]

For the last few months i have been testing the Agisoft Photoscan's 4D timeline tools. Unfortunately Photoscan has a bug when doing the alignment step, it can't handle more than 200-300 frames, anything above that and it crashes. This is independent on alignment quality, i can set it to lowest quality and it still crashes. I have tried it across multiple projects with different data but it still crashes. Also, it would be great if we at least could do the alignment of only a few frames, which is not possible, it aligns all the frames which means that with this bug Photoscan is as good as useless for 4D photogrammetry. I have send many crash reports but i have yet to receive a report on what the issue is. I will upload a recording of the alignment and the crashing for you guys to see, maybe it helps to figure out the problem.

https://www.youtube.com/watch?v=VVub_A1QTM0&feature=youtu.be

[Alexey Pasumansky]

Hello Meije,

Have you added any comment or your e-mail address to the crash reports, so that we can idenfity them?

We haven't been reported such issues before, even though I've seen users that are processing much larger 4D sets.

[Alexey Pasumansky]

Hello Meije,

Can you do the following:
- run image matching operation from Python console using matchPhotos() function with the parameters that you are using, for example:

Code: [Select]

PhotoScan.app.document.chunk.matchPhotos(accuracy = PhotoScan.Accuracy.HighAccuracy, preselection = PhotoScan.Preselection.NoPreselection, keypoint_limit = 40000, tiepoint_limit = 4000)- after operation is finished, save project in PSZ format and send it to us, so that we could try to reproduce the issue during the alignment stage on our side.

Also have you checked the memory consumption during the process?

[Meije]

I have found a workaround!

So the issue appears to be that after the alignment process the updating of the 3D view port causes Agisoft to crash. So this is what i did, i clicked on a camera from the list, this then shows me the photograph in the view port instead of the 3D view port. I then run the alignment and everything works!,  as it does not have to refresh the 3D view after alignment (which would cause it to crash).  So there is a 3D viewport refresh issue (maybe it momentarily tries to load all frames into the view port?). Anyhow, this workaround worked on all quality alignment reconstructions.

[Meije]

I hope this helps, let me know if you need any more info. I tried this on 4 machines on my university and all had the same problem (with different data sets), so it's definitely a bug.

[Alexey Pasumansky]

Hello Meije,

Can you send the project in PSZ format with the alignment results to support@agisoft.com, so that we could try to reproduce the issue on our side, running camera alignment basing on the matching points present in the project? Original photos are not required.