Abstract: Acquiring labeled data for the training a classifier is very difficult, times consuming and expensive in the area of remote sensing. Many semi-supervised techniques have been developed and explored for the classification of remote sensing images with limited number of labeled samples. In this paper, a new unsupervised clustering algorithm is proposed by combining [k]-means and bee colony algorithm. Features of remote sensing images are extracted by Gray Level Co-occurrence Matrix（GLCM） and wavelet transform, and then [k]-means clustering of feature dataset is performed. The initial clustering center is generated by the maximum-minimum product-neighborhood averaging method. The new swarm algorithm and [k]-means algorithm are alternately implemented to achieve remote sensing image clustering. With the comparison experiment of the UCI dataset and the Liangshui National Nature Reserve remote sensing image data, the algorithm has high clustering accuracy and meets the application requirements of remote sensing image clustering.

Key words: remote sensing, [k]-means clustering, bee colony algorithm

摘要： 在遥感领域，获取用于训练的标记数据耗费巨大且困难，因此许多非监督技术逐渐被发展和应用于标记样本有限的遥感图像。将[k]均值和蜂群算法相结合，提出一种新的非监督聚类算法。使用灰度共生矩阵和小波变换提取遥感图像特征，对特征数据集进行蜂群[k]-means聚类。整个聚类过程首先使用最大最小距离积邻域均值法产生初始聚类中心，将蜂群算法和[k]-means算法交替执行，实现遥感图像的聚类。通过UCI数据集和凉水国家级自然保护区的遥感数据的实验结果表明，该算法具有较高的聚类准确率，满足遥感图像聚类的应用需求。

关键词: 遥感图像, [k]-means聚类, 蜂群算法