Research On Algorithm Of Morphology Classification Of Neuronal

Brain has the most complex structure and the powerful function in vivo, which contains more than hundreds of billions of nerve cells (neurons). Human Brain Project (Human Brain Project, HBP) is designed to study database of neural information all over the world to establish common standards by multi-disciplinary integration analysis of large amounts of data. It can help us speed up the understanding of the brain.It's our problem how to design neural pattern recognition and classification methods according to databases of neuroinformatics. To solve the problem, we need to do in-depth analysis of the samples of neurons. When picking up most'effective features'of neurons, we can make a comparison between sample of neurons and'new neuron'. Then we can judge which class the'new neuron' belongs to. Based on the theory, this paper builds the pattern recognition model of neurons through template matching method and support vector machine method. Then the neurons can be classified.Template matching method in pattern recognition is the simplest method. When given a certain amount of samples, the method can offer a good identification results to us. Support Vector Machine (SVM) has developing rapidly in the decade of the 20th century. It's the core content of the statistical learning theory. In this paper, we use template matching method and support vector machine method to make the morphological classification of neurons. We can see support vector machine method has good classification ability through making a comparison between the above methods. Sometimes, we also find that it's not ideal to use feature vectors which have different dimensions to design a classifier in the study. In order to obtain correct classification, it is necessary to make data pre-processing. The result shows that:the data pre-processing plays a very key role in the classification on geometric feature.