模糊C均值聚类的波段修正法变化检测  

Change detection of spectrum correction method based on the fuzzy C-means clustering

作　　者：姜莹[1] 万幼川[1] 李刚[1] 

机构地区：[1]武汉大学遥感信息工程学院,武汉430079

出　　处：《测绘科学》2015年第5期74-79,共6页Science of Surveying and Mapping

基　　金：国家863计划项目(2013AA122104-3);高等学校博士点科研基金课题(20130141130003)

摘　　要：针对遥感影像变化检测中同物异谱、异物同谱导致的波段间敏感性差异、孤立噪声干扰等问题研究的不足,文章提出了一种基于模糊C均值聚类的波段修正法变化检测算法:采用差值/比值复合法构造分波段对比差异影像,增强了复合影像的振幅及结构信息;基于差异影像的邻域熵权修正多波段联合的邻域互信息量建立修正影像,较好整合了影像不同波段间的反射差异和邻域空间信息;最后利用改进的模糊C均值算法对修正影像进行变化检测。多组实验结果表明,波段修正法缓解了单波段敏感性差异对变化检测的影响,有效避免了局部最优、孤立噪声干扰等问题,检测结果更接近客观实际。Aiming at the insufficiency of studying the sensitivity-difference between the bands and the interference of the isolated noise due to the same object with different spectrum or different object with same spectrum, the paper proposed a change detection algorithm of spectrum correction based on fuzzy C- means clustering., a compound algorithm of difference and ratio was used to structure the sub-band con- trast image in order to enhance the amplitude and structure information of composite image~ a corrected image was established by using neighbour-entropy of difference image to correct multi-band neighbour mu- tual-information, for effectively integrating the reflection-difference between the bands and neighbour spa- tial-information; finally, an improved fuzzy C-means was used to realize change detection. Experimental results showed that the proposed method could alleviate the influence of the sensitivity difference of single band on the change detection, and avoid the local optimum and noise interference effectively, making the results more closer to the real situation.

关 键 词：变化检测 模糊C均值聚类 波段修正法 分波段对比 多波段联合 邻域熵权 互信息量 相似度加权法 

分 类 号：TP75[自动化与计算机技术—检测技术与自动化装置]