Dongba classical ancient books image classification method based on ReN-Softplus convolution residual neural network

Dongba culture is a treasure in the world's historical civilization. In the traditional method of classification and preservation of Dongba ancient books, it is mainly through the translation of the scriptures of Dongba priests, and the artificial identification of damaged books is carried out, and then a large amount of manpower is spent on each The book is compiled and compiled, and finally classified. This paper proposes a Dongba classic ancient book intelligent image classification method based on ReN-Softplus convolution residual neural network: in the structural design of convolutional neural network oriented to Dongba image, by using the principle of residual network, and using 1*1, 3*3 and 5*5, which are relatively small convolution kernels, construct a network suitable for Dongba image recognition and classification, and then form a final network structure by repeated superposition. In addition, combining the smoothing characteristics of the Softplus function with the advantages of the better sparse representation of the ReLUs function, a ReN-Softplus activation function is proposed to improve the performance and classification accuracy. Applying this method to the classification identification of Dongba classic ancient books, the test results show that the accuracy of the AlexNet method is increased by 0.7%, and the time of the VGGNet method is reduced by 18s.