论文部分内容阅读
针对模糊C均值(Fuzzy C-Means,FCM)算法,不能有效地对夹杂噪声的遥感图像聚类的问题,本文提出了一种基于局部空间信息核模糊C均值(Kernel Fuzzy C-Means,KFCM)的遥感图像聚类算法。首先,运用核函数将遥感图像的所有像元映射到高维特征空间,通过非线性映射优化遥感图像的有用特征;然后,根据相邻像元之间的相关性,利用一种空间函数重新定义像元的模糊隶属度,将像元的局部空间信息引入到FCM算法中,并在高维特征空间中使用这种基于局部空间信息的FCM算法对像元聚类。由于引入了像元的局部空间信息,算法可以直接应用于原始遥感图像,不需要滤波预处理。大量实验结果表明,本文提出的基于局部空间信息KFCM的遥感图像聚类算法具有较强的抗噪能力,可得到较好的同质区域,优于现有的FCM算法、模糊局部信息C均值(Fuzzy Local Information C-Means,FLICM)算法及KFCM算法。
Aiming at the problem that fuzzy C-Means (FCM) algorithm can not effectively cluster remote sensing images with noisy noise, this paper proposes a Kernel Fuzzy C-Means (KFCM) method based on local spatial information, Remote Sensing Image Clustering Algorithm. Firstly, kernel function is used to map all the pixels of the remote sensing image into the high-dimensional feature space, and then the useful features of the remote sensing image are optimized through the non-linear mapping. Then, according to the correlation between adjacent pixels, a spatial function is redefined The fuzzy membership of pixels is used to introduce the local spatial information of pixels into the FCM algorithm, and the FCM algorithm based on local spatial information is used to cluster the pixels in the high dimensional feature space. Due to the introduction of local spatial information of pixels, the algorithm can be directly applied to the original remote sensing image without filtering pretreatment. Numerous experimental results show that the remote sensing image clustering algorithm based on local spatial information KFCM proposed in this paper has strong anti-noise ability and better homogeneous region, which is better than the existing FCM algorithm. The fuzzy local information C means ( Fuzzy Local Information C-Means, FLICM) algorithm and KFCM algorithm.