Using the trajectory from visual-inertial odometry (VIO)

Hello,

We operate in GPS-denied environments, and our data-acquisition robot records a video. Using VIO, we could obtain the robot trajectory, and use it to add a geolocation to the images.

This could enable the use of the “Use Triangulation of Image Geolocation” and/or “Use Distance” matching strategies.

With these more advanced matching strategies, I hope to reduce the processing time and make the reconstruction more robust (thanks to loop closure).

One issue however: the trajectory obtained with VIO is often drifting. So the images would not be geolocated very precisely at a large scale. Therefore, I would like Pix4D to use the geolocation only for the matching strategy. In the optimisation process, it should give a very low weight to these positions, because they are probably inacurate.

Is that possible? How to set up Pix4D for this case?

I would love to hear from anyone who has experience in combining VIO and photogrammetry!

Thanks,

Nicolas

Hello Nicolas,

The way to give more or less weight to the image geotags is by typing the value in the Image Properties Editor:

By default, the software takes 5-10 m but you can adjust it as you want.

The higher the value is, the less weight the geotags will have and vice-versa.

Regards.

 

 

 

Thanks Daniel for your reply.

Just to be sure: does the geotags accuracy has any influence on the selection of image pairs for matching?

 

Hello again,

The software first finds keypoints in the images and then the matching is done.

The accuracy that you set up for the geotags should not have any effect in the image pairs selection for matching.

Regards.

Excellent, thanks a lot for your answer!