Research On The Recognition Mechanism Of Visual Nerve Face Recognition And Its Application

As an important method of biometric authentication,face recognition is a vital technology of our society,and it is a hot research field of pattern recognition and image processing.Most of Traditional facial recognition algorithms achieve face detection or locating with a priori knowledge of face and facial features,and then fully dig out the facial features or gradient information,and finally face recognition is implemented by the use of template matching or depth learning strategies.Compared with human visual system,such a mathematic model of face recognition may leave a huge deviation with biometric visual neural mechanism.So,we introduce a dynamic neuron model with sensitivity and peripheral suppression characteristics based on Leak Integrate-and-Fire neuron model in this paper.Besides,we constructed visual neural pathways to fulfill rapid and effective face recognition with visual perception.There are several biometric visual mechanisms we used in the visual neural pathways,including visual receptive fields,direction selection,selective attention,and neural coding of visual information flow,etc.The research work and achievements of the thesis include the followings:(1)A new face recognition method based on biometric visual information flow and processing mechanism was proposed.DOG receptive field and Gabor multi-directional response unit were constructed to simulate the retinal cells and neuron ganglion cells to filter the visual information.The LIF neuron was used to code the direction selection feature of the visual path,and the feature data of the face can be effectively extracted and selected.(2)A new method of face salience information extraction based on selective attention mechanism was proposed.The local entropy was calculated on the face image,and it was used to generate neuron spike to locate the position of feature enrichment in the image.At last,the direction selection features were enhanced to the salient image.Experimental results showed that the selection of attention mechanism can effectively locate the information-rich region of the input face image,and also suppressed the interference of peripheral noise,and improved the accuracy of face recognition.(3)A novel face recognition system based on dynamic neuron(DLIF)model was proposed.Considering the important role of the complex connections between dendrites and axons in visual information processing and transmission,neuron sensitivity and peripheral inhibition were used to dynamically recognize thedistribution of information in the input region,and stimulation of an outer range of signal source would inhibit the spiking of DLIF neurons.Global information fusion changes the idea of using local information in face recognition.The experimental result showed that our method's accurate was superior to the classical PCA face recognition method's,the accurate of it was 95.63%,which showed that the optimized system had better face recognition ability.