摘要
在运用于遥感图像的分类时,为考虑图像像元间的空间相关性,首先在聚类的迭代过程中根据相邻像元的隶属度,确定邻域内的优势类别,同时引入反映空间相邻关系的加权系数,修正中心像元的隶属度。其次考虑算法用于图像分割的通信复杂度及动态聚类时的空间相邻关系,提出了相应的并行实现方案。最后,通过实验数据证明了算法在减少聚类的迭代次数以及提高聚类结果精度等方面的有效性,其并行方案也取得了较好的线性加速比。
Considering the spatial relationship of pixels when it is used in classification for remote-sensing imagery, Neighbor-based FCM algorithm was put forward by modifying the value of fuzzy membership degree with the neighbor information during the clustering iterations. We use dominant class, if it can be determined in a fixed neighbor region, or the weighted parameters based on the distance of neighbors to perfect the membership degree of central pixel. Then parallel implement for the algorithm was also proposed by taking account of the communication complexity and the spatial relationship for image partition. In the end, the experimental data indicate the efficieieney of the algorithm in decreasing the clustering iterations and increasing the classified precision, and the parallel algorithm also achieves the satisfying linear speedup.
出处
《计算机应用》
CSCD
北大核心
2007年第10期2512-2514,2517,共4页
journal of Computer Applications
基金
武器装备预研项目(403050203)
关键词
模糊C均值聚类算法
邻域信息
模糊隶属度
并行算法
Fuzzy C-Means （FCM） clustering
neighbor information
fuzzy membership degree
parallel algorithm
