Clustering Recognition of Neurons Based on Multiple Image Features

Abstract

A new method combined multiple image feature extraction ways with self-organizing feature map neural network is presented for clustering recognition of neurons. Firstly, the features of gray histogram and gray level co-occurrence matrix of neuron images are computed respectively, these two kinds of features are combined as the feature vectors of various neuron images. Then the feature vectors are clustered by self-organizing feature map neural network, the similar neuron images are grouped into same area. There are 160 image samples for the testing finally, and the accuracy of clustering recognition is up to 91.3%. The experimental results show that the new method has the higher identification accuracy than a single image feature extraction method for clustering recognition of neurons, which can be used to identify the neurons in the preliminary study.