Multiplexed illumination for classifying visually similar objects

Distinguishing visually similar objects like forged/authentic bills and healthy/unhealthy plants is beyond the capabilities of even most sophisticated classifiers. We propose use multiplexed illumination to extend range that can be successfully classified. construct a compact RGB-IR light stage images samples under different combinations illuminant position colour. then develop methodology for selecting patterns training classifier using resulting imagery. model synthetically relight samples, greedy pattern selection scheme exploits this ability train in simulation. apply trained carry out fast classification new objects. demonstrate approach on artificial real fruit showing marked improvement compared with fixed-illuminant approaches as well more conventional code scheme. This work allows previously indistinguishable objects, potential applications forgery detection, quality control agriculture manufacturing, skin lesion classification.