Spiking Neural Networks And Its Application In Image Segmentation

The Spiking neural network is called the third generation artificial neural network,it can simulate the dynamic firing characteristics of biological neurons and transmit information by using the precise timing of pulse firing.It has been proven by several research-teams that it could deal with clustering and classification problems related to time patterns.Its time coding characteristics make the processing of information,neuron model and synaptic learning rules of the network very different from those of the previous two generations of artificial neural networks.Doing a research on its modeling and mechanism and its application in image segmentation has theoretical significance and practical application value.In this paper,an image segmentation method based on Spiking-Self Organizing Feature Map(Spiking-SOM)neural network clustering was designed and implemented.The experimental results showed that this method has some potential application value.First,we carried out the experiment of image denoising,and the median filtering method was selected to improve the image quality.Then,the super-pixel was calculated,the Spiking-SOM neural network was constructed by using IF(Integrate-and-Fire)neuron,and the weight matrix of Spiking-SOM was constructed based on the color information of super-pixels to realize network weight matrix initialization.The target clustering of super-pixel images is realized by using Hebbian rule to train the network.Finally,the image segmentation is realized based on the clustering results.The mainly works of this paper is as follows:1.To make the theoretical basis for the research methods,the structure,coding and application of Spiking neuron model and Spiking Neural Networks were investigated in this paper.2.This paper also discussed the influence of STDP(Spike Timing Dependent Plasticity learning rules on the population-distribution behavior of Spiking Neural Networks?tested the effect of the main parameters of Spiking Neural Networks on the number of Polychronization Groups during STDP training and explained the phenomenon of spatio-temporal locking by Polymorphic synchronization Group.3.An image segmentation method based on Spiking-SOM neural network clustering was designed and implemented.First,the target image was preprocessed;Second,calculated the superpixel and constructed the Spiking-SOM neural network,besides,the network weight matrix was initialized and the Hebbian rules training network is adopted.Last,realized this method simulation and made the comparative evaluation with other methods.