Fascinating, I had no idea that flag was copyrighted. I guess that's one of the reasons why there was a push for alternative flag designs a few years ago.

As far as searching for prior art, I would probably look at using a large data-set like the [1] 2.8 million art images released by the Smithsonian. So you can process it all locally rather than uses a SAAS api.

Then use something like tensorflow to do the image recognition. Here's a [2] fun tutorial.

[1] https://www.smithsonianmag.com/smithsonian-institution/smith...

[2] https://codelabs.developers.google.com/codelabs/tensorflow-f...