Image classification method based on visual cortex processing mechanism and pulse supervised learning

Abstract

The invention discloses an image classification method based on a visual cortex processing mechanism and pulse supervised learning. The image classification method based on a visual cortex processingmechanism and pulse supervised learning includes the steps: 1) inputting an image in a mode of dynamic image, and then processing the input under three different space-time scales; 2) solving responseof a V1 layer simple type Neuron; 3) solving response of a V1 layer complex type Neuron; 4) solving response of a V4 layer Neuron; 5) training output layer connection; and 6) inputting a test sample.The image classification process of the image classification method based on a visual cortex processing mechanism and pulse supervised learning is closer to the processing process of a real brain, and the image classification method can preferably extract the local orientation information of the input image so as to preferably perform image classification. Besides, the image classification only needs training connection of the output layer without training layer by layer, thus having higher efficiency. For the image classification method, classification accuracy on a hand-written digital setis around 96%, and the classification accuracy is high.

Description

technical field [0001] The invention relates to the technical field of image processing, in particular to an image classification method. Background technique [0002] On the problem of image classification using neural networks, there are currently two types of networks. One is the traditional neural network, such as BP neural network and convolutional neural network. Although they have high accuracy, they need to use error backpropagation to adjust the connection layer by layer, resulting in relatively low learning efficiency, and using There is a big difference between neurons and biological neurons, so they lack biological support. The other is the spiking neural network (SNN), which is closer to the biological reality because it uses spiking neurons (accumulate the input, and the neuron will have an output after reaching a certain level). There are relatively few methods for processing image classification in SNN, including liquid state machines and STDP-based classif...

AI Technical Summary

Problems solved by technology

[0003] In view of this, the object of the present invention is to provide a kind of image classification method based on visual cortex processing mechanism and pulse supervised learning, it preprocesses image by simulating the processing mechanism of brain visual cortex, and then uses pulse supervised learning to adjust output Layer connection, so that the classification process is closer to the processing process of the real brain, and at the same time solve the problems of low efficiency and low classification accuracy of the pulse neural network when using traditional neural networks to deal with image classification problems

Images