Abstract: In view of the shortcomings of the existing intuitionistic fuzzy C-means clustering, nonlinear function is adopted to map data samples from Euclidean space to high dimensional feature space of Hilbert, and kernel space intuitionistic fuzzy clustering algorithm is gotten. At the same time, taking account the interaction of neighboring pixels, the neighborhood pixels are integrated into objective function optimization of kernel space intuitionistic fuzzy clustering algorithm, and kernel space intuitionistic fuzzy clustering segmentation with pixels local information is obtained by mathematical deduction. The test results of graph segmentation show that kernel space intuitionistic fuzzy C-means clustering algorithm is more satisfactory in segmentation results compared with the existing intuitionistic fuzzy C-means clustering segmentation method, and the kernel space intuitionistic fuzzy C-means segmentation method with local information is shown to be more robust.

Key words: intuitionistic fuzzy clustering, kernel space, local information

摘要： 针对现有直觉模糊C-均值聚类仅适合呈团状数据的不足，采用非线性函数将数据样本从欧式空间映射至再生希尔伯特高维特征空间，得到核空间直觉模糊聚类算法;同时考虑相邻像素的相互影响，将邻域像素融入核空间直觉模糊聚类的最优化目标函数中，经数学推导便得到嵌入像素局部信息的核空间直觉模糊聚类分割算法。图像分割测试结果表明，核直觉模糊C-均值聚类分割法相比现有直觉模糊C-均值聚类分割法能获得更满意的分割效果；同时，嵌入局部信息的核直觉模糊C-均值聚类分割法表现出良好的抗噪鲁棒性。

关键词: 直觉模糊聚类, 核空间, 局部信息