一种利用花粉算法优化的遥感图像分割方法

为提高遥感图像分割的准确性与抗噪性,以学生t分布混合模型为基础,结合K-means与花粉算法的特点,将K-means算法局部寻优能力强以及花粉算法全局寻优能力强的优点相结合,提出一种基于K-means的学生t分布混合模型,用于遥感图像分割。该方法中,根据学生t分布与高斯分布以及柯西分布比较接近的特点,对花粉算法的执行过程进行改进。将K-means算法与改进后的花粉算法结合以提高聚类效果,从而快速确定混合模型参数初始值。混合模型最终参数的求解主要采用EM算法,以提高算法最终的图像分割效果。仿真图像和实际图像实验结果显示,该研究得到了比对比方法分割精度更高、稳定性更好的分割结果。

A Remote Sensing Image Segmentation Method Optimized Using Flower Pollination Algorithm

To improve the accuracy and noise immunity of remote sensing image segmentation,a student’s t distribution mixture model(TMM)based on K-means is proposed.Based on the TMM and the characteristics of K-means and flower pollination algorithm(FPA),the proposed algorithm combines the advantages of strong local optimization ability of K-means algorithm and strong global optimization ability of FPA.In this method,the implementation of the FPA is improved based on the fact that the student’s t distribution is close to the Gaussian and Cauchy distributions.The K-means algorithm is combined with the improved FPA to improve the clustering effect,thereby quickly determining the initial value of the mixture model parameters.The final parameters of the mixture model are mainly solved by the EM algorithm to improve the final image segmentation effect of the algorithm.The experimental results of the simulation image and the actual image show that the accuracy of the proposed method is higher than that of the contrast methods,and its stability is better than that of the contrast ones.