Research On Recognition Method Of Calligraphy Stylein Deep Belief Network Driven By Multi-Volume Layer

The Calligraphy is the national treasure of China,and it also plays a particular role in the world arts.The study of calligraphy style based on iconology has significant meaning of Calligraphy recognition,classification storage,retrieval and style appreciation etc.In recent years,with the wide application of image processing and recognition technology in the field of text recognition and other related fields,the traditional image feature extraction algorithm has the defects that the image features can not be correctly expressed and the recognition rate is low.In this thesis,deep learning is introduced to conduct an in-depth study of calligraphy style recognition.A calligraphic style recognition model based on deep belief network(DBN),convolutional neural network(CNN)and convolutional neural network and restricted Boltzmann machine is designed.The main work of this thesis is as follows:1.In calligraphy style recognition based on deep belief network,a deep belief network model with three hidden layers is proposed to solve the problems of low recognition rate and weak generalization ability of the deep belief network model with few hidden layers.Firstly,after the original calligraphy image is cut in a single word and preprocessed by word,feature extraction is carried out on the hidden layer;and then feature fusion is carried out in high order hidden layer;Finally,the fused feature vector was input into the Softmax classifier to complete the classification.The experiment results show that the network model has a good recognition effect on the data sets of five standard Windows calligraphy and four calligraphy datasets.2.In the calligraphy style recognition based on convolutional neural networks,the network structure based on LeNet-5 model and the corresponding training methods are studied.On the basis of the design and optimization of the model,the method of setting the convolution kernel size,the type of activation function,the pooling method,the size of the pooling kernel and the learning rate are systematically studied,a new convolutional neural network model is proposed.The final experiment results shows that,the proposed network model can be effectively used for calligraphy style recognition and has a good recognition effect on two kinds of calligraphy style data sets.3.In calligraphy style recognition based on convolutional neural network and restricted Boltzmann machine,a hybrid network model combining convolutional neural network and constrained Boltzmann machine is studied and proposed.Firstly,feature extraction and dimensionality reduction were carried out on the input calligraphy image through convolutional neural network;and then the feature vector was input to the restricted Boltzmann machine for deeper feature extraction;Finally,Softmax classifier was used to complete the classification.The experiment results shows that the hybrid network model has better recognition effect in calligraphy style recognition compared with deep belief network and convolutional neural network model.