Vertical Misalignment with local control GCP's

[RHoney]

Using the DJI Phantom 3 Professional for stockpile surveys at our mine site. Captured 180 images of an area of about 200m x 100m. Set out 6 GCP's with RTK GPS and imported into Agisoft.

The model aligns with our previous GPS surveys pretty well spot on, in the horizontal plane. However we are getting an error in the vertical alignment. The southern end of the model is ~1.5m low and the North is ~1.5m high, the centre is good. The GCP are spread over the surveyed area and confident in there accuracy. When comparing the model and the GCP's in a separate software the 1.5m discrepancy between the model and the GCP's is present. The errors in the markers are 0.7-1m for all of them. I have manually selected the markers in each photo which didn't help the errors.

I created a model for a second area and ran into the same problem.

How can I rectify this?

Regards

[Alexey Pasumansky]

Hello RHoney,

Are you unchecking all the cameras in the Reference pane before performing marker-based optimization?

[Mdaemon]

Hello,

I am usually using high accuracy GPS RTK (Trimble) for the GCPs. I am using no GPS information for the camera positions.
LAst time I recognized, that the height information is getting inaccurate to the edge of the model.

Are there any recommendations, how wide the gcps should be spreaded to the egde of the measurement?

Are there any rates of bulging of the model?

Regards

Martin

[SAV]

Hi Martin,

At the edges of the model the image overlap is generally poorer, hence the error of check points/GCPs increases in these regions.
Note: The area that is surveyed by UAV/drone should generally be larger than what you want to use in your final 3D photogrammetric model. This will help you to avoid the before mentioned boundary issues.

I suggest to generate a processing report in Photoscan to see the image overlap at the edges of your model and use it to delete regions with low overlap/high error from your final model.

Cheers
SAV

[Mdaemon]

Hi Sav,

i am doing that report. Is it possible to export the overlapping map to qgis?
I am still meaning the bulging of the model outside of the gcps when overlapping is still ok.
any ideas on that?

regards
Martin

[mwittnebel]

It would be nice to export a geotiff with the overlap.
Could you provide the Report ?

[SAV]

Hi Martin,

Unfortunately there is no easy or convenient way to get the overlap map into qGIS or any other external software package. #featurerequest

You can try it the 'dirty' way. Save the map from the PDF report as png/jpg and then manually overlay and georeference it in qGIS.

@Alexey: Would be nice if one had the option to export the overlap map as geoTIFF. Moreover, it would be very useful if one could specify to crop the point cloud, DEM or orthophoto based on image overlap. Or do you already have a Python Script for that which you are happy to share with the community? 

Regards,
SAV

[Alexey Pasumansky]

Hello SAV,

I think with the next version we would be able to create the script that will allow to produce the image footprints as shape files in faster and more convenient way than the script that is looping through all the polygons of the model.

Basing on that information it should be possible, as I assume, to get the overlap information in the GIS applications and probably we could also extend the script to get some overview information fo the overlap. But yet the required Python API functionality about to come in 1.3.1.