The current integration tests form a solid foundation for wider automation of the regression tests. Use it.
And create instructions for the customer how they can use it for their safety app testing.
Think how to deal with the reference image database, that might grow in size. And examine the approach used in the Lancelot system.
Basically the current approach could also be used in conjunction with the capacity test easing curve testing quite easily, if the assumed large reference data size wouldn't be an issue.