Cognitive Foundations and Formal Theories of Human and Robot Visions

The neurological and cognitive mechanisms of visual information representation and manipulations in the visual cortex are fundamentally different from those of knowledge in the long-term memory. It is recognized that visual information is elicited and retained as feature frames rather than individual bitmap images. This keynote lecture presents the latest development of the Cognitive Vision Theory (CVT) that extends and synergizes the classical vision principles based on Dennis Gabor's wavelength and orientation filters (1946) and David R. Rubel's hypercolumn theory (1981, Nobel Prize Laureate in Physiology or Medicine). This talk formally explains the visual information processing mechanisms in the brain. A Spike Frequency Modulation (SFM) theory is introduced to coherently link the pixel-based CCD mechanism of the retina to the Hubel feature representation in visual cortex throughout the entire neural pathways of human vision. Experiments on the CVT and SFM theories in both human and robotic visual information processing are rigorously elaborated. The basic studies on the CVT and SFM theories provide a new perspective for human and robot vision, which pave a way to novel image processing applications in AI, neural networks, image recognitions, sequence learning, computational intelligence, unmanned systems and robot vision.