Research On Image Semantic Recognition Of Helan Mountain Rock Paintings

The research content in this paper is the image semantic recognition of Helan Mountain rock paintings.The research on image semantic recognition has become a hot issue in the field of pattern recognition and image processing,artificial intelligence and computer vision,and it has occupied an increasingly important position in the field of image analysis.Image semantic recognition expresses the content information contained in the image in language,inputs an image,and outputs the semantic information corresponding to the image,so that the computer has the same function as the human brain to describe the things seen by the vision.Through the study of the research on image semantic recognition by scholars in recent years,the work done in this paper is mainly divided into the following points:(1)Making image data sets of Helan Mountain rock paintings.Helan Mountain Rock Painting Scenic Spot is located in Yanhua Road,Helan County,Yinchuan City,Ningxia Hui Autonomous Region.Through on the spot investigation,some rock painting pictures were collected using photographic equipment and after the later processing,making an image data set of Helanshan rock paintings that meet the experimental needs.(2)Aiming at the low extraction efficiency and low accuracy of traditional image feature extraction methods,this paper chooses to use the convolutional neural network algorithm to extract low-level features of images,and studies the structure,advantages and disadvantages of convolutional neural networks.(3)Aiming at the high computational complexity of convolutional neural network image recognition algorithm based on direct support vector machine,this paper proposes a convolutional neural network based on kernel function binary tree support vector machine,which will be based on Gaussian kernel function and binary tree structure support vector machine.The convolutional neural network combines to generate a new model,which reduces the computational complexity and improves the recognition accuracy.(4)For the classical convolutional neural network model,there are many parameters,and it is easy to produce over-fitting.As the number of network layers deepens,the more parameters need to be trained,the higher the computational complexity.And the training model for different data sets,the recognition rate needs to be improved.In this paper,an improved activation function and a convolutional neural network model based on the cuckoo search algorithm are proposed.The activation function of the classical convolutional neural network is improved.For the training of model parameters,the cuckoo search algorithm is used to find the optimal parameters.Finally,the feasibility of the improved algorithm is verified by experiments.