Multi Level Neural Information Interaction Model And Its Application

The human visual perception system can recognize,detect,analyze and understand the target object in a very short time,and its efficiency and accuracy are difficult to achieve in the current computer system.It is of great significance to study the mechanism of visual perception and apply it to image recognition and detection.Considering the close relationship between the visual mechanism and the efficiency of biological visual perception,this paper constructs a multi-level neural information interaction model and applies it to image processing.Firstly,a response model based on visual fine perception mechanism is constructed by introducing the classical receptive field orientation selectivity,non classical receptive field per ipheral inhibition and high-level visual cortex feedback mechanism.Then,the fusion of color feature information is considered to enrich the extraction of image information.The neural information interaction is realized by using the cross receptive field coding mechanism,which simulates the subjective initiative of visual perception,strengthens the continuity of edge information,and completes the detection of image edge information.Finally,the parallel processing channel of primary visual cortex visual pathway is introduced to integrate local information and overall information,and the edge detection is realized by combining the fine visual processing of advanced visual cortex.The innovation and research results of this paper are as follows:(1)Considering the hierarchy and feedforward of visual information transmission,a contour detection method based on the combination of azimuth hierarchical thinning and texture dynamic suppression is proposed.Firstly,the fine perception method of visual featu res of classical receptive field is proposed to improve the detection accuracy of contour and orientation;in addition,multi-scale feature fusion strategy is adopted to realize multi-scale optimization of orientation response for the previous optimal orie ntation;then,a dynamic suppression mechanism of non classical receptive field response is simulated,which can effectively remove the texture details in the response;finally,the vision is constructed The final contour response is obtained by the feedba ck fusion model of saliency information.The results show that the optimal performance index is 0.48,which is better than other mainstream methods.(2)Considering the coding mechanism of cross receptive field and the initiative of visual saliency information,an edge detection method based on the fusion of neural information interaction and dynamic optimization of visual saliency information is proposed.Firstly,the vision information fusion based on color control is proposed to enrich the color image information;secondly,the mechanism that the cross receptive field discharge activity can encode more precise vision information is used to realize the fine interaction of vision information;the neuron peripheral excitation fusion mechanism is used to ret ain the detail information to highlight the difference between the edge information and the background;finally,the subjective visual perception is considered Initiative,build a multi-directional dynamic optimization model of visual salient information,strengthen the continuity of edge information,and complete the efficient detection of image edge information.Taking the edge detection of color colony image with rich details and overlapping adhesion as an example,the average value of reconstruction similarity index is as high as 0.87,and the detection effect is significant.(3)An edge detection method based on visual path parallel processing channel is proposed.Considering the color characteristics of the image,the color components are fused in resp onse.The multi-path parallel processing channel is introduced into the primary visual cortex.On the one hand,the local information is integrated based on the small area of the primary visual cortex.On the other hand,the contour saliency map is formed based on the saliency map principle of the primary visual cortex.The local information is well preserved while the irrelevant information is removed.The high-level visual cortex coding fine visual information mechanism is introduced to form the overall contour,and the feedback fusion is used to strengthen the detail information to realize the image edge information detection,and is applied to the road lane line detection.The detection effect is obviously better than the comparison method,which shows the feasibility of this method.