VIO "Q" (quality) metric and distant features

[hmlow]

Hi devs

Can i trouble you to shed some light on the following:

1. How is the quality metric in the qvio server computed? I noticed that when features are distant (outdoors, maybe more than 100m away) the Q metric drops to -1 even though the features are plenty (>20 feature points) and well spreaded out across the image frame. Im on SDK 0.9.5

2. Is there an upper limit to the feature point distance (amidst having plenty of features) when the VIO algo starts becoming unreliable?

3. When my VIO camera is looking at distant features (as described above), sometimes it starts drifting, and some other times it continues to work well. This is an expected correct behavior?

Thanks

[Chad Sweet]

The code where quality is computed can be found here: https://gitlab.com/voxl-public/voxl-sdk/services/voxl-qvio-server/-/blob/master/server/quality.h