基于模糊推理的最大似然分类算法研究

本文将模糊集理论与最大似然分类原理相结合,用模糊均值和模糊协方差代替传统最大似然分类的均值和协方差矩阵,依据极大隶属度原则,对最大似然分类算法进行改进。并尝试采用一种基于相异像素空间分布算法的分类精度评定方法,得到相异像素空间分布图,依据该空间分布图对不同的分类方法进行精度评定。实验结果表明,改进后的最大似然分类法的正确率、Kappa系数均优于传统的最大似然分类方法,所采用的精度评定方法也较传统方法在有效性、效率等方面有所改善。

Algorithm research of maximum likelihood classification based on fuzzy inference

This paper integrates the principle of Fuzzy Set with the principle of Maximum Likelihood Classification(MLC),and substitutes the traditional mean value and covariance matrix in MLC using the fuzzy mean value and fuzzy covariance.Then it improves the MLC algorithm based on the principle of maximum membershi Pdegree.Furthermore,this paper adopts one kind of classification precision evaluation methods based on the algorithm of space distribution of different classified pixels,and obtains the space distribution ma Pof different classified pixels.It also carries on the precision evaluation of classification based on this map.The experimental results indicate that the total classification precision and the Kappa coefficient of the improved MLC surpass the precision of traditional MLC method,and the new precision evaluation method is better than the traditional methods in the respects of validity,efficiency and so on.