Visual Positioning System for Navigation

Dear guys, we cannot as yet find a system that combines drone forward and or downward camera vision and analysis with the google earth images. It should not be difficult to make virtual beacons (bridge, road, hill tops, towers) on the forward camera and then compare with google earth data of these points and then bring the camera view in that and by triangulation find the raw camera position of the drone? Such system could be used over quite some distance eg few miles if big objects are used as virtual beacons. With And are there people working on it?

Pls can anyone bring us a bit further in this?

Cheers, Winfried

We had some discussion related to this topic here:

Basically we are talking about a TERCOM inspired navigation system.

Thanks, highly appreciated. Think that a TERCOM system is even more sophisticated and can be used with good preparation and over long distances. What we hope is that there is a AI system that can identify suitable virtual beacons (road crossings, hills, mountains, etc…) in google maps at take off position of the drone, and then compare those with the forward camera view at that height. Such comparison would calibrate the drone position… Then when the drone flies it will generate different forward images of moved beacons. The google earth virtual view can than be matched and thus position be notified? Presume that there are many ways and experts in this, is there a solution… Cheers W

If you look at the link I provided, it refers to

A seasonally invariant deep transform for visual terrain-relative navigation

I tried on simulator a few month ago and I was inspired by this paper (below is presentation):

Yes this is OK for pattern recognition in different light conditions (can that also be done with google earth pro? and we will go through. The vertical cam is OK but might be giving only a smaller view (of course depending on camera height) Horizontal might enable more territory coverage. I was quite inspired by this one: The beacons making AI etc… Cheers, W