Abstract
Light fields in the real world are highly complex because they contain multiple illumination sources, including indirect light reflected from other surfaces. Such interreflexions are well compensated by the human visual system, by employing prior knowledge about the nature of the light field (Mach-card experiment: Bloj et al., 1999). Interreflections are the rule in natural environments and therefore a general problem for all organisms relying on colour vision, e.g. bees. Honeybees have excellent colour constancy as demonstrated for flat Mondrian-like scenes (Werner et al., 1988). However, bees have far fewer neurons available for these operations than humans and it is therefore interesting to ask: how can a complex colour vision task, like compensating the effect of light fields in 3D scenes, be solved by a relatively simple system? I will present a Mach-card experiment with free-flying honeybees (apis mellifera), where individual bees were trained to discriminate between green/white cards folded in different spatial configurations: (1) planar (2) concave and (3) convex. The stimuli (paper cards, 2.5 x 2.5 x 6 cm) were presented in front of a vertical, "grey" background (a turntable disc, 80 cm diameter). Previous experiments had shown that bees are able to resolve the depth profile of similar 3D forms. The results show that the bees recognized a trained green/white pattern irrespective of its spatial configuration, in other words, bees compensated the occurring interreflexions. Further experiments, involving simulations of the interreflexions, showed that the compensation depends on the consistency of spatial and chromatic cues. This suggests that bees use their knowledge about the spatial configuration in 3D scenes in order to discount the effects of mutual illuminations. Whether this involves cognitive inference, as it suggested for human colour vision, or is driven by hard-wired interconnections, remains to be solved.
Meeting abstract presented at VSS 2014