On the basis of the original capsule network, the pre-convolution structure is firstly introduced to extract high-level features of the mural image; secondly, the fitting performance of the model is enhanced by adding uniform layer activation; finally, the adaptability of the capsule network is enhanced, on the basis of improving the gradient smoothness Optimization using adaptive learning rate improves the classification accuracy of the model. An art style painting classification algorithm based on information entropy is proposed. First, seven representative landscape art painting styles are selected as the research objects, and the images are preprocessed such as denoising and normalization. Secondly, extract the style characteristics of painting artworks, obtain the color entropy, block entropy and contour entropy of the image respectively, and combine them to form the information entropy of different painting styles. When the information entropy is obtained, the color space is converted into the Lab color space, and the color entropy of the image is obtained through the color values of the a and b channels and the weighting function; the information entropy of the block is obtained by dividing the information entropy of the landscape art image The mean value is used to obtain the block entropy; through the Contourlet transformation, the contour information of the landscape art image is obtained to obtain the contour entropy. Next, combine the extracted color entropy, block entropy, and contour entropy, and use support vector machine (SVM) to learn and train artistic style images to obtain a classification model of artistic painting style; SVM classification recognition obtains the final classification result.