基于高斯云变换的遥感图像多粒度聚类-计算机科学.pdfVIP

Vo1. 44 No. 9 第 44 卷第9 期 计算机科学 2017 年 9 月 COMPUTER SCIENCE Sep.2017 基于高斯云变换的遥感图像多粒度聚类 刘旋王国1m，罗小波 (重庆邮电大学计算智能重庆市重点实验室 重庆 400065) 摘 要 遥感图像技术的迅猛发展，使得传统聚类方法的局限性日益凸显。针对其信息量大、结构复杂等特点，从多 粒皮、多层次的角度来分析与理解地学现象，能够更好地解决遥感图像的自适应聚类问题。基于云模型与混合高斯相 结合的高斯云变换是一种求解多粒度问题的新方法，能够解决问题域中多粒度的生成问题，但是其时间复杂度较高以 及对嗓声敏感等缺点，导致对遥感图像的聚类结果不理想。因此提出一种改进的高斯云变换方法，首先通过K-Means 聚类优化初始粒度的选择，其次结合幅度云综合对粒度跃升策略进行改进，然后使用一种隶属度距离进行粒度的区域 划分，最终对遥感图像进行聚类。实验结果验证了所提方法的正确性和有效性。 关键词 遥感图像，高斯云变换，多粒度，图像聚类 中固法分类号 TP391 文献标识码 A DOI 10. 11896/j. issn. 1002-137X. 2017. 09. 004 Multi-granularity Clustering of Remote Sensing Image ßased on Gaussian Cloud Transformation LIU Xuan WANG Guσyin LUO Xiacγbo (Chongqing Key Laboratory of Computational Intelligence.Chongqing University of Posts and Telecommunications.Chongqing 400065.China) Abstract With the development of remote sensing image technology , the limitation of the traditional image analysis methods have become increasingly prominent. From the perspective of multi-granularity and multi-level ,we can solve the adaptive clustering problem of remote sensing images better ,with large amount of information and complex struc tures. Gaussian cloud transformation which is based on cloud model and Gaussian mixture model is a new model of multi-granularity method. It can extract multiple concepts from different granularities in a problem domain. However , due to its time complexity and noise sensitivity ,the clustering result of remote sensing images is not ideal. An improved Gaussian cloud transformation method was proposed in this thesis