Given two photo datasets of the same location taken at different times (or even without two data sets) it would seem that there would be an ability to remove shadows as an additional feature that would appeal to your user base. Is this something that has been discussed or is under investigation? (Note: I'm not talking about removing the drone shadow - I'm talking about removing all shadows.)
https://www.cs.sfu.ca/~mark/ftp/Eccv02/shadowless.pdf
http://static.googleusercontent.com/media/research.google.com/en//pubs/archive/37743.pdf
http://research.google.com/pubs/pub37743.html
The problem with this is that almost everywhere in the world that is not right on the equator the shadows will always fall to one side of any vertical object. It seems like a cool idea but requires two data sets that have lighting coming from different enough angles to be able to mutually fill in the dark zones.
Please sign in to leave a comment.
According to the papers, it seems shadow removal will work anywhere, at any time of day depending on the contrasts of the objects. There are some pretty interesting models in the Google paper as well as the SFU paper. Also, in the most diminutive case that doesn't use the sophisticated one-pass shadow removal referenced it would not be onerous to make two passes of the same area at different times of the day that created sufficient shadow differences.
If one of the methods is used (shadowless.pdf) it may be the case that it even produces a new revenue item - a 10'x10' (estimated) mat of reference colors to include in every survey to get correct profile on the camera being used. I'd pay $200 for that if I could remove shadows from surveys. I'm led to believe, however, from the Google paper that it is not 100% necessary to even use that method.
"depending on the contrasts of the objects" is the key phrase here. This is a very researchy topic that isn't widely employed because of the limitations.
If you can provide two data sets and show us that it works we will look at adding it to our system.
