Reality Capture uses only 45 of the 327 images available

Development Asset Creation
,
1

I am processing a survey performed with UAS SenseFly Ebee, but the software uses only 45 of the 327 images available to create a rough model. Any suggestions to solve the problem? Thank you

2

Hi Walter,
what kind of land have you photographed? Is it forest one? If so, there could be issues with alignment and in this case (forest) it could help to flight in higher height.

Also, what was the overlap between images? Are you using those images as georeferenced?

3

Salve, ho sorvolato un’area del fondovalle in parte agricola e in parte antropica. Le immagini sono georeferenziate in RTK. La sovrapposizione è di circa il 75%.

4

Hard to say without seeing the images. The RTK coordinates are baked to images? Or do you have created flightlog with the image coordinates? What were your alignment settings?

5

The RTK coordinates are linked to the images and not taken from the flight log. The alignment settings are the standard software ones.

6

Is it possible to share your images with us?

7

Sure. At the link you can find the images:

Thank you

8

Can you upload the images to a folder I sent you a link for? Thanks.

9

Sure. They are loading right now.

10

Thank you for your images.

I was able to align around 200 images with these alignment settings:

As you mentioned RTK I increased the camera priors positions accuracy.

Problem in your dataset is the height of the flight, it seems quite low to me. Also, there is insufficient overlap between images. When we were using eBee, we flight in doblegrid for such low heights (it seems like 50 - 60 m to me). Also it seems, that it was windy during flight, as there is a lot of tilted captures (not just on turns).

Or it could be a way how the images were captured. If so, I would use also perpendicular images + tilted.

11

I apologize if I open this discussion again, but a colleague was able to align the images with ****************** and create the orthophoto. Actually I did it too with ********** by changing the SFM algorithm. There are some small holes, but they are acceptable.

12

Hi Walter, how many images were aligned there? RealityCapture uses different algorithms for alignment therefore it is more sensitive to the inputs and how they are captured.