为提高分类精度,提出一种基于最大期望(EM)与遗传(GA)算法的多尺度SAR图像无监督分类方法。利用多尺度自回归(MAR)模型描述SAR图像中不同尺度之间的统计相依性,提取多尺度特征。应用混合模型描述多尺度特征,并将GA算法与EM算法相结合给出混合模型的参数估计算法,利用最小描述长度(MDL)准则选择模型的分量数。最后使用Bayes分类器实现了图像的分类与分割。该方法集EM算法和GA算法结合后的优点,对设定初值有较少的敏感性,因而避免了局部最优解。应用于SAR图像的实验表明,在分割精度上GA-EM方法优于MAR模型的算法。
An efficient unsupervised multiscale classification of synthetic aperture radar(SAR) imagery based on the genetic algorithm with expectation maximization(GA-EM) algorithm is proposed.This algorithm is capable of selecting the number of classification of SAR image using the minimum description length(MDL) criterion for Gaussian mixture model.Our approach benefits from the properties of genetic algorithms(GA) and the EM algorithm by combination of both into a single procedure.Therefore,our algorithm enables escaping from local optimal solutions since the algorithm becomes less sensitive to its initialization.Some experiment results on the SAR images are given based on our proposed approach,and compared with EM algorithms.The experiments show that the GA-EM outperforms the other method on segmentation precision.