Alignment Inconsistencies

[ruehilton]

Hi everyone,

I hope you're well.

I am getting inconsistent alignments. 

Model 1
The first model I created used only my high level drone photographs (105 cameras). I was happy with this model though it misses a lot of the detail from the front elevation of the building: https://skfb.ly/6nP7s

Model 2
I thought, no problem, I will now run a model with all of my pictures (ground and drone) to fill in the gaps - 303 cameras total. This model however came out worse - and I suspect that there are too many cameras which has confused the software?? https://skfb.ly/6nSCB

Model 3
I thought, no problem, I will re-run a model with the drone pictures with the addition of the ground level photos which show only the missing elevations of the building - to fill in the gaps.

This model has captured the areas missing in the second model, but completely missed the elevations of the building which were fantastic on models 1 and 2.

Obviously I stopped following the build of dense cloud. Each model is taking many hours to produce.

Can anyone offer help or an explanation please?

Thanks,

Rue

[ruehilton]

Apologies, I meant cmeras not chunks - I have modified the above.

Rue

[Alexey Pasumansky]

Hello Rue,

The issues may be related to the different conditions when the images from different sub-sets were taken and also to the difference in the viewing position. Can you share the project data with support@agisoft.com?

[ruehilton]

Thanks Alexey. I have emailed the data as requested.

Rue

[Alexey Pasumansky]

Hello Rue,

I think I have finally figured out the source of the issue. You have added NEF files to the project, but PhotoScan doesn't support Nikon RAW format, instead of that the preview images 160x120 pixel were added, so it's expected that such low resolution images cannot be aligned.

I can suggest to convert the images to TIFF format, for example, and then reprocess together with the DJI data. Also I've noticed that the EXIF information for drone images is lost, it may happen if you were preprocessing the images in the external image editor or used video frames, so I also recommend to input known focal length and sensor pixel size (both in mm) to the corresponding fields of the Tools Menu -> Camera Calibration window prior to processing according to the camera specification. (sensor pixel size is usually about 0.004 mm, so it shouldn't be confused with the image resolution in pixels).