Research And Application Of K-means Clustering Algorithm

Data mining technology is a crossover subject which colligates database,machine learning,AI and other fields.Data mining can dig out the information from the disorder,clutter and a large number of data.It has been achieved a mass of theories and methods,themain research concentratre on the clustering which is based on the distance,for instance K-means clustering is the most classical algorithm.K-means algorithm can be considered as the most important unsupervised machine learning approach in clustering.It is a partition clustering algorithm,all the data is divided into k sub-classes which are quite different,Through such the iterative partitioning,k-means algorithm minimizes the sum of distance from each data to its clusters.Because of easy implementation and efficiency,it is popular and widely used in many fields,such as data compression,image segmentation,machine learning,abnormal data analysis and statistical disciplines.However,some limitations still exist.For example,if the selected initial cluster centers are not suitable,it is easy to fall into local optimal solutions and could not guarantee stable results.In this paper,in-depth study and analysis of the clustering algorithm in the K-means clustering algorithm,summed up its strengths and weaknesses.Considering the characteristics of the k-means algorithm,in this paper,the K-means algorithm is applied to the tracking technology of video object.As to the independence of the k-means to the initial centers selection,we present two new initial centers selection algorithms,and use a large number of experiments to verificate the impact of the randomly selected initial value to the clustering results.The researches and contributions are as follows:(1)In this paper,the K-means algorithm is applied to the tracking technology of video object.Firstly,a sample model is established based on the background pixels of some video image to simulate the relevant action characteristics of the object.Then,we can use the clustering algorithm for dividing the sample model,and detect the background pixel of the image according to the sample model.By the way,the sample model is updated in using the class of the classification of the related pixel in the sample model,so as to further improve the detection efficiency of video image background.(2)This paper proposes a new improved k-means algorithm by using the fast local convergence of mean shift and the characteristics of regional division,it can reduce the overall number of iterations and the complexity of the algorithm.It can also enhance the global convergence and the stability of the algorithm.Through the experimental data,the new improved algorithm can enhance the stability of the results,and improve the accuracy of the clustering results.