Research On Image Clustering Based On K-medoids Algorithm

In this highly information-intensive society,digital image information is also rapidly increasing,how to effectively process and inquire massive image data and obtain the potential information from it is of great importance.Therefore,image data mining also came into being,becoming a hot topic of scholars in recent years.The process of extracting potentially useful knowledge from a large amount of data is Data mining,its main research content includes cluster analysis.Cluster analysis uses unsupervised learning methods to classify data with similar characteristics into one class,so that the similarity between data in the same class is high,and the similarity between data in the classes is low.Image clustering is based on the image as the target data for clustering,so that high similarity images are classified into a category,thereby enhancing the image management and retrieval performance.As a traditional clustering algorithm,K-medoids clustering algorithm is widely used because of its advantages of simple principle,high implementation efficiency and insensitivity to "noise".However,the K-medoids clustering algorithm needs to determine the K value and the initial centroid in advance,and is not easy to jump out of the local optimum.Based on this,many scholars combine K-medoids clustering algorithm with swarm intelligence bionic algorithm to optimize the performance of K-medoids clustering algorithm.In this paper,improved cat swarm optimization algorithm firstly,then the improved cat swarm algorithm is combined with the K-medoids clustering algorithm,so that the clustering algorithm of K-medoids clustering can be clustered under the specified centroid.Finally,the application of K-medoids algorithm in image clustering is used.The main work is as follows:(1)This paper introduces the basic concepts,algorithms flow,characteristic of K-medoids clustering algorithm.It introduces two modes of cat swarm optimization algorithm and it's characteristic.And simply describe the basic theory of image clustering.(2)Improved cat swarm optimization algorithm's Velocity-Displacement Formula in Tracking Mode.In this paper,the cat swarm is approached to the optimal solution by increasing the inertia weight and related factors,At the same time,in order to avoid "precocious" and improve the convergence of the algorithm,the constant mr is replaced by the linear mixing ratio based allocation formula,which makes the cat swarm algorithm tend to the global optimum.(3)The improved cat swarm algorithm is combined with K-medoids algorithm to solve the problem that K-medoids algorithm is not easy to jump out of the local optimum and chooses the initial centroid randomly.(4)Preprocessing the image,including color image graying,image feature extraction and.image enhancement,the improved algorithm proposed in this paper is applied to image clustering.The experiment proves the feasibility of the algorithm.Experiments show that the improved cat swarm optimization algorithm not only has good stability and good convergence,but also has a higher accuracy when clustering datasets with K-medoids clustering algorithm.Finally,the experiments on the images also get better clustering results.It proves the feasibility of the algorithm and its applicability in the image.