摘要
针对Fisher线性判别法和传统的Bayes判别方法在遥感影像聚类问题研究中存在的不足,提出一种以隶属度代替先验概率的模糊Bayes-Gauss聚类算法,并将此算法应用于真彩色(RGB)图像中的草地、道路、裸土地和建筑物的聚类.实验结果表明,本算法在聚类中与Fisher线性判别法和传统Bayes判别法相比,具有精确度较高、误识率和拒识率较低、适用性较强的特点.
In face of the shortcomings of the Fisher linear discriminant method and the traditional Bayesian algorithm in remote sensing image clustering,a new kind of fuzzy Bayes-Gauss clustering method was proposed by using fuzzy membership functions instead of priori probability.And the new algorithm was applied to the clustering of grasses,bare land,roads and buildings in True Color（RGB）.The results of experiments showed that the new algorithm for clustering had higher precision,lower false accept rate and false reject rate,and stronger applicability advantages,compared with the weighted Fisher linear discriminant and traditional Bayes algorithm.
出处
《集美大学学报（自然科学版）》
CAS
2011年第2期154-158,共5页
Journal of Jimei University：Natural Science
基金
福建省自然科学基金资助项目(A0810014)
福建省教育厅重点项目(JK2009017)
福建省科技厅青年人才创新基金项目(2009J05009)
厦门市科技计划项目(3502Z20093018)
