Analysis And Research Of Image Feature Extraction Model Based On Hierarchical Method

Image recognition and classification technology as one of the research hotspots of artificial intelligence has important research significance.Image feature extraction is the key of image recognition technology,which determines the accuracy of image recognition in essence.The traditional way of image feature extraction often depends on specific extraction method designed for a specific problem,although has made great achievements,but it`s adaptability is often not strong,so the automatically extracting of image features become the focus of the present task in image recognition.With the advent of big data era,the feature learning method represented by depth learning model shows its powerful ability in image feature extraction task.By the way of layer by layer training,the depth learning model can extract the deep features of the image in a highly autonomous way,and get the best results in the existing research stage.Although the learning ability of deep learning model is extremely powerful,the complexity of its model structure and the overhead of training also hinder the training.The specific work is as follows:(1)The common clustering algorithm K-mean algorithm is introduced briefly,and its advantages and disadvantages are analyzed in depth.Based on the K-mean algorithm,the spherical K mean algorithm is introduced,and an improved algorithm is proposed to improve the clustering effect.(2)On the basis of the improved spherical K mean algorithm,the clustering algorithm is applied to the image feature extraction,and the task of image classification is trained by constructing a two-layer learning model.Experiments show that the two-layer model proposed in this paper has better effect than single layer clustering method.(3)The basic principle of convolutional neural network is deeply analyzed,and its training process and model structure are studied.Based on Deep Brief Neural Networks and the relevant technology will use convolutional neural network,combining the depth of convolution neural network has been built and used to model training,and achieved good results.