Study On The Model And Several Working Mechanism Of LGN In The Bionic Brain

Bionic visual system is a very sophisticated image processing technology. In recent years, although digital image technology has gained considerable development, it's still not perfect in some visual tasks. The organism, however, has a relatively more powerful visual system, so maybe the establishment of visual information processing system based on the biomimetic will be able to make up for the shortage of digital image technology.The main object of this paper is the lateral geniculate nucleus (LGN) in the biological visual system. By tracking and in-depth analysis of modern neuroscience research findings, first, this paper discusses the basic structure and several working mechanism of the lateral geniculate nucleus, and accordingly establishes the structure and working model of the bionic LGN, and then builds up the software of the bionic LGN. At last, this paper demonstrates the mathematical connections contained in the LGN.1. Physiologically, there are two main cell types in the primate LGN:relay cell and interneuron. The former is responsible for processing and transmitting information received from the retina, and the interneuron regulates the information transfer process by suppressing the former. Relay cells can be divided into three channels:K, P and M channel. They transfer the blue/yellow antagonistic information, the red/green antagonistic information and the luminance information respectively to achieve the parallel and independent visual information delivery. The interneuron inhibits the relay cells to enhance the visual contrast of the image. Based on these theories, this paper builds up the bionic LGN which is organized in columnar structure containing all cell types in the biological LGN. Every column in the bionic LGN contains 22 different cells totally, which are responsible for different aspects of binocular information processing. The distribution of the cell column and the receptive field structure of the cell in each column are designed according to the characteristic of the biological LGN. 2. The receptive field structure of LGN relay cells has a specific mathematical meaning. A retinal ganglion cell provides excitatory input to several LGN relay cells at the same time, and there exists a homologous category between these relay cells; While a relay cell receives several retinal ganglion cells in its receptive field, and there exists a same-target category which means these retinal ganglion cells project to the same target relay cell. There are same-target categories simultaneously in the ON and OFF channel of the LGN. The categories between these two channels are able to be switched to each other by center-off functor and center-on functor separately. This means that when the external stimulus changes, the pathway transferring this stimulus changes correspondingly, indicating that the LGN has a good adaptability to the external light stimulation.3. When the LGN relay cells provide information to the primary visual cortex (V1), it also receives feedback signals from the simple cells in the sixth layer of V1. The orientation selective simple cells have specific feedback connections with LGN relay cells, and provide excitatory signals both to the relay cells which lie in the simple cell's receptive field center and corresponding interneuron. The result of feedback makes the response of the relay cells which locate in the border of the image to be strengthened, and then the corresponding response of simple cells will also be strengthened. This dynamic feedback regulation of visual system can improve the sensitivity of the information on the boundary. So the bionic LGN will use the same feedback regulation system.4. The feedback regulation system of the LGN makes some specific relay cells' response to be strengthened. As these relay cells strengthened by the same one simple cell, there exits a resonant category between these relay cells which means that they act like they resonate. The set of resonant categories of all relay cell types constitutes the boundary contour of light stimuli. On the contrary, in the lateral inhibition system of interneuron, some relay cells are suppressed by the same one interneuron, there exits a co-inhibitory category between these relay cells. The relationship between relay cells ultimately makes the gap between the responses of neurons further widen, particularly on the boundaries of the image. The interaction of the closed-loop feedback mechanism and the lateral inhibition mechanism in the LGN form the neurophysiologic basis in the boundary identification of visual system. The software of LGN is written based on the model of bionic LGN, and experiments about the basic biological function of LGN are performed. Experimental results show that:1) the size of relay cell's receptive field increases with the distance to the fovea. This variation is conducive to the attention mechanism of the visual system, and forms the foundation of the attention mechanism.2)Software LGN relay cells show a specific pattern in response to stripe stimulation, and different types of relay cells have a specific enhanced reaction on the boundary of image, which forms the basis of boundary extraction function of bionic vision system.3) The lateral inhibition between the LGN relay cells is very suitable to improve the image contrast, and thereby improves the clarity of the image and highlights the image edge information.4) Positive feedback effect of simple cells is able to enhance responses of LGN relay cells which lie on the border of image, making the image boundary information strengthened further in the LGN level. In general, the image information is processed by the receptive field of LGN relay cells, effected by the lateral inhibition, and then strengthened by the signal of cortical feedback, and the final result is the image boundary elements are highlighted which lay the foundation for object recognition.