应用光学, 2012, 33 (4): 766, 网络出版: 2012-09-12
改进的K均值聚类红外目标检测方法
IR target detection based on improved K-means clustering
摘要
利用图像方差能很好地反映目标边缘信息的特点,提出一种基于方差的K均值聚类红外目标检测算法。利用形态学方法对红外图像进行预处理,运用相应的模板计算得到红外图像的方差图像,利用K均值聚类算法对方差图像进行聚类,从而分离出目标类别和背景类别。实验表明,该算法提取的红外图像中目标信息的兰德指数最高,说明该算法能有效地提取红外图像中目标信息,从而达到目标检测的目的。
Abstract
Considering the variance of image was a very good response for edge information, a target detection algorithm by K-means clustering algorithm based on variance was presented. First, this paper prepressed the infrared image by morphological method, and calculated the corresponding variance image by using a specific template, then gathered each difference image class by using the K-means clustering method, finally the different target edge information was got. Experimental results show that the algorithm can effectively extract the IR target edge.
姜斌, 石峰, 崔东旭, 张鹏辉, 袁轶慧, 张俊举. 改进的K均值聚类红外目标检测方法[J]. 应用光学, 2012, 33(4): 766. JIANG Bin, SHI Feng, CUI Dong-xu, ZHANG Peng-hui, YUAN Yi-hui, ZHANG Jun-ju. IR target detection based on improved K-means clustering[J]. Journal of Applied Optics, 2012, 33(4): 766.