Research On Intuitionistic Fuzzy Clustering Algorithm And Its Application In Gray Image Segmentation

Clustering is a very important unsupervised learning method.By calculating the similarity between samples,it can divide similar samples into the same cluster and dissimilar samples into different clusters.It has the advantages of simplicity and efficiency in data mining tasks.As a generalization of fuzzy clustering algorithm,intuitionistic fuzzy clustering algorithm can more clearly describe the fuzzy information in the objective world,so it is widely used in various fields.However,the algorithm still has many problems,such as the initial membership matrix has a great influence on the clustering results,and is sensitive to noise and outliers.Therefore,this paper studies the above problems of the algorithm,and puts forward the corresponding improved algorithm,the specific work is as follows:Firstly,in order to solve the problem that the intuitionistic fuzzy clustering algorithm generates the initial centers randomly,which leads to fall into the local optimum easily.An intuitionistic fuzzy clustering algorithm based on degree centrality is proposed.By setting a threshold,the sample data is constructed into a network,and the reciprocal of the distance between the sample data is taken as the edge weight in the network.The method of finding important nodes by using the degree centrality of nodes in the network is used to calculate the degree centrality of nodes in the network.The nodes with large center and unconnected in the network are selected as the initial clustering centers,and the corresponding membership matrix is calculated.Experimental results show that compared with the intuitionistic fuzzy clustering algorithm,the improved algorithm improves the convergence speed and stability of the algorithm.Secondly,when dealing with the data with uneven distribution of samples or noise,the membership of intuitionistic fuzzy clustering algorithm is very large for some data due to the constraint of membership,which affects the clustering effect.In this paper,a relaxed membership constrained intuitionistic fuzzy clustering(RM?IFCM)method is proposed algorithm relaxes the membership constraint of one sample to that of all samples,so as to reduce the impact of data far away from the cluster center on the cluster center.Aiming at the problem of too large or too small membership in the algorithm,a new membership correction method is proposed.Compared with the possibility fuzzy clustering algorithm and the suppression fuzzy clustering algorithm,it is found that the anti-noise ability of the algorithm is stronger.Thirdly,when RM?IFCM algorithm is applied to gray image segmentation,it is found that the algorithm cannot use the spatial information of the image,and it is difficult to achieve the desired effect for the image with complex gray distribution and significant noise.Therefore,this paper proposes RM?IFCM algorithm based on weighted pixel distance,which uses the mean membership degree of neighborhood pixels to weight the distance from pixel to cluster center,so that the membership degree of noise data is closer to that of neighborhood pixels,so as to reduce the impact of image noise on segmentation results.Experimental results show that the RM?IFCM algorithm based on weighted pixel distance is better than other algorithms in the segmentation of gray image with noise.