摘要
提出了一种针对光学遥感图像海域分割的自适应期望最大算法(Expectation Maximization,EM)。传统EM算法需要给出高斯混合模型的个数,文中所做的改进使其具有自适应确定高斯模型个数的能力,由此可以降低由于个数选择的不合理带来的错误分割的风险。将该方法应用低分辨率复杂背景遥感图像的海域分割,取得了较理想结果。
An adaptive expectation maximization（EM） algorithm is proposed for sea area segmentation of optical remote sensing images.The number of Gaussian mixture models should be known first in traditional EM algorithm.The improvement in this paper provides the algorithm with the ability of adaptively computing the number of Gaussian models and thus reduces the risk of wrong segmentation due to improper model number selection.Experimental results show the effectiveness of the algorithm in sea area segmentation for low-resolution remote sensing images with complex background.
出处
《无线电工程》
2011年第4期20-22,28,共4页
Radio Engineering
关键词
遥感
期望最大
图像分割
remote sensing
image segmentation
expectation maximization
