Lighting invariant background subtraction for new object detection? Choquet integrals?

I’m trying to detect new objects in a scene that frequently has large lighting changes.

Choquet integrals looked like a promising technique:
El Baf F, Bouwmans T, Vachon B. Foreground detection using the choquet integral. In: WIAMIS’08 Proceedings of the 2008 ninth international workshop on image analysis for multimedia interactive services. New York: IEEE; 2008. p. 187–90.

Is there an implementation for images in openCV / numpy?
Or can recommend a good alternative?

definitely not so ;(

Thanks. Any recommendations for foreground/background classification in video that’s robust to large light changes?

I’d expect Mixture of Gaussian models to do well.

OpenCV has background segmentation/subtraction in two modules: