层次混合模型快速遥感影像分割算法

Hierarchical Gaussian mixture model for fast remote sensing image segmentation

针对高分辨率遥感影像中同物异谱和同谱异物导致的分割困难问题,提出一种层次高斯混合模型(HGMM)快速遥感影像分割算法.首先采用HGMM构建影像的统计模型,其具有准确建模像素强度统计分布呈现的非对称、重尾和多峰等复杂特性的能力;然后根据贝叶斯理论构建基于HGMM的分割模型,为了简化参数求解并提高算法效率,定义均值和方差为关于权重的函数;最后采用共轭梯度(CGM)求解模型参数.实验中采用所提出算法和传统统计模型分割算法分别对合成、全色和彩色高分辨率遥感影像进行分割实验.实验结果表明,所提出的HGMM具有准确建模复杂统计分布的能力,且能够准确和有效地分割全色和彩色遥感影像.

As the same object with different spectrum and the different objects with same spectra of high resolution remote sensing image segmentation lead to seqmentation diffculty, a fast remote sensing image segmentation algorithm is proposed by hierarchical Gaussian mixture model(HGMM). Firstly, the HGMM is used to build statistical model of image, which has the ability to model the asymmetrical, heavy-tailed and multimodal distribution of pixel intensities.Then, the HGMM-based segmentation model can be built by following Bayesian theorem. To simplify the complexity of estimating parameters and improve the efficiency, the mean and variance are defined as weights function. Finally, its model parameters can be solved by conjugate gradient method(CGM). The tests can be done with synthetic, panchromatic and color images by using the proposed algorithm and the methods based on traditional statistic model. The results show that the proposed algorithm can obtain accurate segmentation results and high efficiency, which has the ability to model the complicated distribution of pixel intensities.