Intersectional Leading Industry Based on Data Mining

Abstract

With the development of digital technology, it is of great significance to realize the functions of accurate classification and quick search of Chinese painting images. Traditional Chinese painting recognition system mainly includes two steps: feature extraction and classification. Feature extraction is mainly based on personal experience. Although Chinese painting can be classified, there are still some problems such as easy loss of detail information and low generalization ability of model. Looking for an automatic and efficient recognition technology of painting is a hot spot in the current and future research. This paper first proposes a method based on deep belief network to classify Chinese painting images. On the one hand, the two-dimensional structure information ignored by deep belief network is used to extract high-order statistical information. On the other hand, convolution operation is applied to the network structure, which can reduce the noise and enhance the expression of the original signal features. At the same time, the maximum probability model and sparse regularization algorithm are introduced into the hidden layer constrained Boltzmann machine structure, so as to realize the reasoning of probability and the weakening of the over complete phenomenon. Finally, the feasibility of this method in Chinese painting image classification is verified by experiments.