Directions of development of intelligent real time video systems

Abstract

Two directions of development of intelligent real time video systems (technical vision systems) are considered in the report. First direction consists in increasing intellectuality of video systems at the cost of development of new information basis and dynamic models of video information perception processes, principles of control reading parameters of video information for reducing redundancy of video sequence representation and for adaptation them to the requirements of concrete task. Second direction is associated with development of new architectures for parallel perception and processing of information directly on a video sensor matrix, that excludes the need of detailed programming of sequential processes of reading, analog-to-digital conversion and processing of video information. Peculiarities of construction, mechanisms of attention and adaptation, specialization of neurons on the retinal layers and wide parallelism of information processing on neuron network, which take place in a human visual analyzer, are used as a prototype for both directions.