Research On Image Classification And Recognition Method Based On Hierarchical Discriminant Regression Tree

With the rapid development of Internet technology,the importance of images in people's daily life is increasing.A large number of images and video data have been applied in many fields such as medical imaging,digital libraries,industrial property rights,and remote sensing systems.Therefore,the need for classification and recognition of images has arisen.Image classification is a research topic with important application value.It is a sub-problem of the classification problem and a combination of Image mining and classification mining technology.There are many methods for classification.Common classification models include decision trees,neural networks,genetic algorithms,and statistical models.Among them,decision tree algorithm is the most widely used due to its small complexity and fast calculation speed.However,in the process of practical application,there are still many deficiencies in the existing decision tree algorithms.For example,slowness when processing high-dimensional image data.Therefore,it is of great theoretical and practical significance to further improve the decision tree and improve its performance to make it more suitable for clustering and regression of highdimensional image data.The main research contents of this article are as follows:(1)Aiming at the problem of high-dimensional data clustering and slow regression,The construction and retrieval process of the HDR tree is studied,and the selection of the cluster center and the process of data dimensionality reduction are analyzed during the clustering process of the HDR tree.Combined with BDPCA algorithm,a new discriminant regression tree algorithm—2DHDR tree algorithm is proposed.2DHDR trees algorithm can process training samples faster,and make the tree building process faster with fewer layers of trees.In addition,this paper studies the incremental HDR algorithm(IHDR),and analyzes its construction and retrieval process.(2)In this paper,two sets of experiments were performed,respectively on twodimensional numerical data obeying Gaussian distribution and high-dimensional image data.Experiments on numerical data can more vividly show the HDR tree building process.The high-dimensional image data is mainly used for HDR and 2DHDR treebuilding process experiments,classification accuracy rate experiments,training and test time experiments.There are also IHDR classification accuracy experiments,training and testing total time experiments.Experiments show that the 2DHDR algorithm proposed in this paper has faster construction and retrieval speed than the traditional HDR tree,and the larger the sample,the more obvious the difference.