Aerial imagery without GPS = upside down model

[travis_b]

We flew a site with the UAV to obtain aerial imagery. Unfortunately the GPS tagging did not work on all photos. We do have GCPs.

We tried to align photos with GPS tagging on some photos but not on others. The result was a model that looked blurred or "doubled".

Thereafter, decided to remove the GPS tags from all photos prior to importing into Agisoft. Align photos with generic and thereafter add GCPs and correct alignment.

However, after the generic alignment, the model appears upside down. When I press shortcut 7 to see from above, the points are actually on top of the camera markers.

We are hoping to generate a 3D model to compare to the prior survey to see how much the landfill has filled since the previous survey.

Questions:
1) What causes the model to be "upside down"?
2) Will this have a significant impact on the accuracy of the final model?

Thanks for your help.

[stihl]

Seems like there's a discrepancy between the GCP's and the model.

Did you change the coordinate system from (the default) local coordinates to the appropriate coordinate system?

From my extensive experience I can say that adding GPS information to the photos adds little value to the project. In theory it should align the images faster, however that's only true if all the GPS coordinates of the photos are correct. If they're not it will likely to mess things up more so than when you're not using any geotagged images.

As for accuracy - since you're using GCP's to place the model on the correct position, these coordinates trump the geotagged images as they're usually measured with a far greater accuracy. So not having geotagged photos will not decrease accuracy *IF* the area has enough GCP's and they're well distributed.
So basically the only upside to adding geotagged photos (if they're all correct) is that it's easier to do the initial alignment with the GCP's as the model is on roughly the correct location.

[Alexey Pasumansky]

Hello travis_b,

Can you post a screenshot of the Source values tab and the model overview (with the markers)? If it is sensitive data, please send the screenshots to support@agisoft.com.

The most common reason of "upside down" model is incorrect order of X and Y coordinates on import.

[travis_b]

Alexey - thanks. I have attached a screenshot of the model overview.

The source values - which tab is this - sorry fairly new to this.

Thanks,

Travis

[travis_b]

Alexey - I believe the Source value tab is included here. Thanks.

[travis_b]

stihl - thanks for the input. I think I may have missed the step to switch coordinates but have been using the WGS84 previously for all the NAD83 based data since NAD83 does not come up as an alternative and they are fairly close.  I had just done a few other jobs with Agisoft and no problems - not sure if that is the issue.

Interesting with the GPS and the GCPs. I only used 4 GCPs on the last job and the Agisoft calculated error is significant. I assume that I should increase GCPs to increase the accuracy - from your experience how many do you generally use?

Thanks,

Travis

[darius]

Hi,

I am developing solution to set up array of virtual ground control points basing on Internet transmitted
diff and RTK corrections.
Please tell me if Agisoft can 3D visualize scanned aerial imaginery, image by image, 3D geolocalized
to verify if such virtual CGP array can work properly.

AltHold mode fails to work properly with fixed wind UAS
so I would like to geotag images either with GPS, timestamp or gimbal control gyro .

I try to develop ad-hoc generated array of virtual GCPs based on Internet RTK radio,
then placing GPS enabled, 3G /GPRS smartphones to brodcast diff and corrections
to ground control station/ mission planner, to be uploaded to autopilot with drone via MAVLink protocol

Please let me know your opinion.

darius
manta103g@gmail.com

[Alexey Pasumansky]

Alexey - I believe the Source value tab is included here. Thanks.

Hello Travis,

The coordinates of GCPs seems to be in WGS84, but it is not set up, so PhotoScan assumes that the values are in meters. So you need to define the coordinate system in the Reference pane settings dialog.

[travis_b]

Hi Alexey - a couple of questions with respect to your response.

1. The reference system used was actually UTM, NAD83, Zone12. But I can't seem to locate this in the choices when defining the coordinate system. So I thought WGS84 would be close enough. How do I get the correct coordinate system noted above?

2. I do believe that the WGS84 was selected in the preferences. Do you see something otherwise?

Thanks for your help.

[Alexey Pasumansky]

NAD83, Zone 12 has EPSG code 6341, so you can search it by the code in the coordinate system selection dialog.