Neural Mechanisms Research Of The Processing Of Primary Shape Information In Bionic Vision System

In the biological visual system, the visual information is transmitted, organizedand processed in accordance with the neural connection, the neuron receptive fieldand the neural mechanisms, respectively. At any time, the visual information isinterpreted explicitly in different cortexes, and form cognitive images in thehigh-level cortex. Therefore the key of establishing a bionic vision system isresearching on the neural connections, the neuron receptive field structure as well asneural mechanisms. The cooperating rule of the multiple neural mechanisms is thebasic requirement for shape and color recognition in the bionic visual system.According to the theory of visual pathways and the signal projection relationshipbetween the visual cortexes, the fundamental structure of the bionic vision systemwas constructed. The visual signal pathway of bionic vision system is composed ofthe M-channels, the K-channels and the P-channel. The M-channel, through whichthe luminance signal is transmitted, has the highest speed. The speed of the K-channeltransmitting blue-yellow opponent signals is slower than that of the M-channel. TheP-channel is responsible for transmitting red-green opponent signals, and the speed isthe slowest. Complex neural connections exist among LGN, V1, V2and V4cortexes,such as transmitting between nonadjacent cortexes. The bionic vision system shows astructure of multiple-loop feedback.The receptive field of most-simple cells, simple cells, complex cells,end-stopped cells, intersection detection cells and rectangular detection cells weredesigned in this work. The receptive field of high-level cells is based on that of thelow-level cells. Therefore, the higher the cortex level is, the more complex imagesthose cells can handle.The shape signal processing in the visual system is described, which follows theprinciple that the macro image is composed of micro basic elements. And then the relation is introduced to record the position relationship between basic elements.Therefore, the basic elements and the relationship are two major keys in the shapesignal processing. The simplest cells recognize pixel-level images. The simple cellswork to form orientation. The intersection detecting cells recognize the feature points.Finally the rectangular detecting cells in the higher level cortex identify the macroimage according to the elements and the relationship between the elements.The neural mechanisms are the basic algorithm of bionic vision system forimage processing. Neural mechanisms can be divided into feed-forward, feedback,coupling and lateral suppression. Feed-forward, that follows the many-to-one andone-on-one topology mapping relation, is the basic mode of visual signaltransmission in the bionic visual system. Feedback behaves in the regulation of thehigh-level cortices on the low-level cortexes. According to the level of the cortexbetween target cells and cells of the sixth layer, the feedback can be subdivided intotop-down process and bottom-up process. The coupling is the basic way for signaltransmission between cells in the same cortex. By coupling, the response of adjacentcells which have the same receptive field can be enhanced. This provides animportant guarantee for the identification of continuous contour. The role of thelateral suppression is opposite to that of coupling, excited cells inhibit thesurrounding cells which have the different receptive field through inter-neurons. Bylateral suppression, the most stable visual information could be retained and with thenoise removed.A mathematical model about the receptive field position and contact of neuralmechanisms is built up using the category theory and the topology. The relationshipbetween receptive fields in cortexes constitute position category to prove the positionstability of the receptive fields. The projection relationship between receptive fieldsbetween cortexes constitute position category to prove the projection stability ofneuron. Neural mechanisms can constitute a homologous category and homo-targetcategory, which proves the universality of the neural mechanisms.By simulation experiment, it is found that the process of image recognition canbe stable only with the collaborative work of multiple neural mechanisms. Any single neural mechanism alone has a limited effect on the identification of images. And theopponency organizational mode of the visual channel can help to complete the imageprocessing in bionic vision system. The research conclusions showed thecharacteristics of the bionic visual system and verified the role of neural mechanism,which played an important role in the establishment of the bionic vision system.