利用高斯混合模型的多光谱图像模糊聚类分割
发布时间：2019-08-19 10:22
【摘要】：针对传统分割算法难以实现高分辨率多光谱图像分割的问题,本文提出一种利用高斯混合模型的多光谱图像模糊聚类分割算法。该算法采用高斯混合模型定义像素对类属的非相似性测度,由于该算法具有高精度拟合数据统计分布能力,故可以有效剔除噪声对分割结果的影响。同时,引入隐马尔科夫随机场(Hidden Markov Random Field,HMRF)定义邻域作用的先验概率,并将其作为各高斯分量权值以及KL(Kullback-Leibler)信息中控制聚类尺度的参数,从而增强了算法对复杂场景遥感图像的鲁棒性,进一步提高了算法的分割精度。对模拟图像和高分辨多光谱图像分割结果进行了定性定量分析。实验结果表明:模拟图像的总精度达96.8%以上。这验证了本文算法在分割高分辨率多光谱图像时具有保留细节信息的能力,而且也证实了算法的有效性和可行性。该算法能够实现高分辨率多光谱图像的精确分割。
[Abstract]:In order to solve the problem that traditional segmentation algorithm is difficult to achieve high resolution multispectral image segmentation, this paper proposes a multi-spectral image fuzzy clustering segmentation algorithm based on Gaussian mixture model. The algorithm uses Gaussian mixture model to define the non-similarity measure of pixel to class. Because the algorithm has the ability of high precision fitting data statistical distribution, it can effectively eliminate the influence of noise on the segmentation results. At the same time, the hidden Markov random field (Hidden Markov Random Field,HMRF) is introduced to define the prior probability of neighborhood action, and it is used as the weight of each Gaussian component and the parameter to control the clustering scale in KL (Kullback-Leibler) information, which enhances the robustness of the algorithm to remote sensing images of complex scenes and further improves the segmentation accuracy of the algorithm. The segmentation results of simulated images and high resolution multispectral images are qualitatively and quantitatively analyzed. The experimental results show that the total accuracy of the simulated image is more than 96.8%. This verifies the ability of the proposed algorithm to preserve detail information in the segmentation of high-resolution multispectral images, and also verifies the effectiveness and feasibility of the algorithm. The algorithm can realize the accurate segmentation of high resolution multispectral images.
【作者单位】： 辽宁工程技术大学测绘与地理科学学院遥感科学与应用研究所;
【基金】：国家自然科学基金(41301479、41271435) 辽宁省自然科学基金(2015020090)
【分类号】：TP391.41
[Abstract]:In order to solve the problem that traditional segmentation algorithm is difficult to achieve high resolution multispectral image segmentation, this paper proposes a multi-spectral image fuzzy clustering segmentation algorithm based on Gaussian mixture model. The algorithm uses Gaussian mixture model to define the non-similarity measure of pixel to class. Because the algorithm has the ability of high precision fitting data statistical distribution, it can effectively eliminate the influence of noise on the segmentation results. At the same time, the hidden Markov random field (Hidden Markov Random Field,HMRF) is introduced to define the prior probability of neighborhood action, and it is used as the weight of each Gaussian component and the parameter to control the clustering scale in KL (Kullback-Leibler) information, which enhances the robustness of the algorithm to remote sensing images of complex scenes and further improves the segmentation accuracy of the algorithm. The segmentation results of simulated images and high resolution multispectral images are qualitatively and quantitatively analyzed. The experimental results show that the total accuracy of the simulated image is more than 96.8%. This verifies the ability of the proposed algorithm to preserve detail information in the segmentation of high-resolution multispectral images, and also verifies the effectiveness and feasibility of the algorithm. The algorithm can realize the accurate segmentation of high resolution multispectral images.
【作者单位】： 辽宁工程技术大学测绘与地理科学学院遥感科学与应用研究所;
【基金】：国家自然科学基金(41301479、41271435) 辽宁省自然科学基金(2015020090)
【分类号】：TP391.41
【参考文献】
相关期刊论文 前10条
1 赵雪梅;李玉;赵泉华;;结合马尔可夫高斯模型的双邻域模糊聚类分割算法[J];计算机辅助设计与图形学学报;2016年04期
2 赵雪梅;李玉;赵泉华;;基于隐马尔可夫高斯随机场模型的模糊聚类高分辨率遥感影像分割算法[J];电子学报;2016年03期
3 蒋璐媛;肖鹏峰;冯学智;李云;朱榴骏;;基于亚分数混淆矩阵的中国典型区大尺度土地覆盖数据集评价[J];遥感技术与应用;2015年02期
4 王醒策;文蕾;武仲科;周明全;田l，
本文编号：2528187
本文链接：https://www.wllwen.com/kejilunwen/ruanjiangongchenglunwen/2528187.html
