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

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