Visual Cognitive Mechanism Of Color And Shape Features Binding And Study Of Computer Modeling Methods

The purpose of this paper is to explore the cerebral neuromechanism and cognitive process of feature abstraction,storage and bind of color and shape,and to research construction method of the corresponding computer models.Image is composed of different dimensions features visual information such as color and shape,in order to identify an object of external world,the brain is obliged to integrate the information from different cerebral areas to form a holonomic consciousness.The research on cerebral cognitive mechanisms and building up computer models has become a hot and difficult point in the field of computer vision.In recent years,scholars did a substantial number of researches on simulating human visual system,however,little on abstraction of lower specific features,storing and binding cognitive mechanisms and researches on computer models.Consequently,it has a great theoretical and practical significance in both researching neuromechanism and cognitive process on visual features binding and simulating visual neuromechanism to structure computer models.This study designed corresponding experimental paradigms,collected functional Magnetic Resonance Imaging(fMRI)data;utilized the statistical parametric mapping method and independent components analysis to analyze the fMRI data,analyzed the active region related to the task about the identification and binding of color and shape;introduced bi-stage genetic algorithm,structured dynamic causal modeling,optimization process is more stable and efficient,optimization result is more excellent;on the basis of the simplified Pulse Coupled Neural Networks(PCNN),proposed a feature binding PCNN modeling based on the vector(V-FBPCNN),achieved the color image sepration and bundled;proposed the feature binding PCNN modeling based on visual characteristics(VC-PCNN),fulfilled the color image separation and resolved the separation and binding problems of the same color in different regions;through analyzing the structures and pathways of visual system,constructed a binding modeling of the visual cortex functional architecture based on the visual feature binding PCNN modeling.The main work and innovation of this paper are as following:1)Designed experimental paradigms,tested the behaviors,and collected the fMRI data.According to the task of color and shape binding,utilized the theory of experimental psychology,designed the cognitive experiment of color and shape feature binding,conducted the behavioral experiment,cooperated with hospitals and collected the fMIR data.2)Analyzed fMRI data,and revealed the active brain regions doing color and shape feature binding.With the statistical parametric mapping method,the theory of visual pathway was proved,providing that the single feature binding is different from the double feature binding,the encephalic region responsible for processing colors is BA37 area,for shape is BA 18 area,for binding is BA37 area,BA7 area and BA6 area and so on.Analyzed and explored the results of different tasks through utilizing BOLD signal and independent components analysis.3)Introduced bi-stage genetic algorithm to structure dynamic causal modeling(DSGA-DCM).This project utilized dynamic causal modeling to study the effective connectivity triggered by different visual stimulations in the brain visual cognitive network.Regarded the visual stimulation activity as the destabilization in designing modeling,evaluated the DCM through Bayesian estimation,selected the optimal design model,obtained the effective connectivity between visual cortexes under the visual stimulation,revealed the color and shape feature of the cerebral visual cognitive areas and the color and shape feature binding cognitive neuromechanism.Introduced bi-stage genetic algorithm to improve the DCM optimization,selected the optimal modeling to uncover the cognitive mechanism of the color of visual cortex and the shape feature binding.4)Proposed V-FBPCNN and VC-FBPCNN.V-FBPCNN extended gray scalar space disposed by PCNN into color vector space,utilized the time matrix to accomplish the feature separation and binding,fulfilled the automatic judgment of iterations simultaneously,structured the automatic binding modeling of color images.Proposed VC-FBPVNN through imitating brain visual feature and using RGB and IHIS space,by importing rollback module and Depth-first search mechanism accomplished color separation,noise processing and the problems of separation and binding the same color in different regions.5)Structured the feature binding modeling of visual cortex functional architecture on the basis of VC-FBPCNN.This paper constructed the feature binding modeling of visual cortex functional architecture on the basis of VC-FBPCNN,based on visual processing theory and the binding model structured by DSGA-DCM.Imitated the mechanism of the visual ventral pathway to structure the modeling that the computers can realize,preliminarily achieved the integration of cognitive science paradigm and neural science paradigm,simulation experiment shows that the modeling has capabilities to materialize the processing of maching and binding object features.