Illumination-invariant landmark representation by color contrasts - Insect Model and Robot Application

点击翻译按钮获取中文摘要

Insects apply information reduction to make sensory input manageable for their tiny brains. For visual navigation it has been suggested that insects use UV-visual contrasts to reduce a complex visual scene to a simple and illumination-invariant skyline representation. The project will validate this hypothesis by constructing a camera-based sensor with UV and visual channels and demonstrating that robust skyline extraction independent of the state of the sky and of the illumination of objects is achieved. Biologically plausible visual navigation models using this reduced input representation will be developed. Both components together will be tested in outdoor environments by simulations on real-world image databases and by robot experiments. The project will, one the one hand, provide strong support for the skyline hypothesis of insect navigation, and will, on the other hand, transfer the simple biological model to robotic applications (robotic lawn mover).