The Research Of Linear Graphics Cognition System Based On Biological Vision Mechanism

The theory and technique of robot visual perception that based on the biological vision characteristics is a leading-edge technology, has already been included in the state's medium and long-term scientific and technological development program. The study requires the integrated use of Visual Physiology, Cognition Neuroscience, Computation Neuroscience, Physiology and Psychology, Robotics, Modern Mathematics and other school subjects'research result. Such research will not only promote the development of robot visual perception theory and technique, but also be favourable for the research of biological vision system's working mechanism and the development of cognition science.According to the requirements of line graphics cognition, This article's main work is the bionic research both on the biological visual system's retina and visual cortex. After fully understanding the biological structure of nerve cells in the retina and visual cortex and their information processing mechanism, this paper puts forward the design of software retina and software visual cortex, and discovers there are some identical mathematical structures in biological vision system and software vision system. These mathematical structures have been given strict mathematical proofs. The main mathematical structures are as follow:1. In fovea, the aggregate of photoreceptor cells that compose visual image and the aggregate of pixels that compose object image constitute the homeomorphous topological spaces. According to the classification of equivalence relations, they constitute the homeomorphous quotient topological spaces. These mathematical structures reflect that the topological relation between the visual image and the object image is consistent.2. The aggregate of nerve cells'response state and the aggregate of receptive fields'information constitute the homeomorphous topological spaces. According to the equivalent classifications, Nerve cells'response state constitute the quotient topological space. These mathematical structures reflect the combinations of nerve cells'state responding to the area of different graphical characteristics. They explain the structure of the image in the form of coded graphic expression.3. The aggregate of receptive fields together with their location relations constitutes category. The category structure reflects the determinate location relationship among receptive fields and the receptive fields in different location mapping different regions'characteristic information.These mathematical structures essentially reflect biological vision system and robot's vision system have the same structural features and information transfer laws. This paper also tries to describe the nerve cells and their receptive fields'structure as well as the intrinsic relations between them with mathematical models, to instruct the design of software visual system and the simulation of the visual information processing.According to the structure of biological vision system and the visual information processing mechanism, and the mathematical framework models which as well as this article proposed , a linear graph cognition system has been established, and has been made experimental research on some visual perception mechanisms and straight line features cognition. Experimental results show that the linear graphics cognition system achieves the following functions :1. Through preliminary established visual perception's initiative attention mechanism and selective attention mechanism in the cognition system, it has been proved that the non-life cognition system also possibly achieves these attentionprocesses. The cognition system can have certain attention psychology ability similar to that of biological vision system.2. Through lateral inhibition mechanism established in software retina which is similar to biological retina, the fact that the lateral inhibition mechanism can be used effectively to highlight line graphics'boundary and contour characteristic, and favour the cognizing of straight line in complicated backgrounds has been proven.3. Through straight line characteristics cognition experiments in simple background and complicated background, it has been proved that the ON/OFF channel mechanism of biology vision system is necessary and feasible to robot's vision perception system. The experimental results also indicate the working mechanisms of ganglion cells,simple cells and complex cells are also necessary and feasible to robot's vision perception system, such as lateral inhibition mechanism,characteristic selective mechanism.4. Through cognition experiments of multiple linear graphics, it has been proved that the cognition system based on the biological vision mechanism can naturally cognize the continuity of straight line and parallel,vertical mutual position relations. The system shows some cognitive processes tallies with the principles of vision psychology and owns some simple psychological abilities that are similar to biological vision system.