一种改进的证据理论C均值分割方法

A SEGMENTATION METHOD BASED ON IMPROVED EVIDENTIAL C-MEANS

证据理论C均值(ECM)作为传统聚类方法的一种改进仍然存在着对噪声敏感和易于陷入局部极小值的缺点,鉴于此,提出一种遗传算法(GA)和证据理论C均值相结合的分割方法,并且在分类过程中引入了位置信息。遗传算法具有全局搜索的能力,很好地克服了证据理论C均值结果局部最优的缺点,而位置信息的引入则解决了对噪声敏感的问题。实验结果证明,该方法收敛速度快,迭代步数少,分割精度高。

Evidential version of the c-means（ECM）,as an improvement of the traditional clustering algorithm,still has disadvantages,such as being sensitive to noise and easily falling into local minimum value.In view of this,a new segmentation method is proposed,which combines the genetic algorithm（GA） and the evidential c-means together and introduces position information into the process of classification.Genetic algorithm has the ability of global searching,so it well overcomes the deficiency of local optimum the evidential c-means has in its result,while the introduction of position information solves the problem of noise sensitivity.Experiment results show that the proposed algorithm has faster convergence speed and less iterative steps and has high segmentation accuracy.