一种改进低层特征算子的遥感影像检索方法  

Remote sensing image retrieval method based on improved Meta-feature

作　　者：周维勋[1] 邵振峰[1] 

机构地区：[1]武汉大学测绘遥感信息工程国家重点实验室,武汉430079

出　　处：《测绘科学》2015年第3期91-95,共5页Science of Surveying and Mapping

基　　金：国家973重点基础研究发展计划资助项目(2010CB731800);国家自然科学基金项目(61172174);国家重大设备专项项目(2012YQ16018505);科技支撑计划项目(2013BAH42F03);教育部新世纪优秀人才基金项目(NCET-12-0426)

摘　　要：在遥感数据检测中,现阶段往往将C_Meta(颜色特征计算得到)和T_Meta(纹理特征计算得到)作为两种不同的Meta-feature特征,使得对于同类地物影像利用C_Meta和T_Meta检索的结果不理想且有较大差异。为了改善检索结果,该文提出通过初步检索选择检索结果较好的C_Meta或T_Meta作为不同类别影像最终的Meta-feature特征;并利用新的影像特征库进行第二次检索。实验结果表明,此方法的检索结果好于原始的C_Meta和T_Meta以及传统的颜色直方图和小波纹理。Meta-feature,one kind feature descriptor,which derives from low-level features,such as color and texture,describes the content of images well and shows good superiority in the field of remote sensing image retrieval.However,C_Meta and T_Meta tend to be treated as two different features in image retrieval.Consequently,retrieval results are not merely unsatisfactory but also with bigger differences for the same image class.With the intention of improving retrieval results,the paper proposed to choose the features with better retrieval results between C_Meta and T_Meta as final Meta-feature of different-class images through initial retrieval process,and then the second retrieval process was carried out with new image features.Results showed that the proposed method had higher precision than traditional methods.

关 键 词：颜色特征 纹理特征 C_Meta T_Meta MF 

分 类 号：TP751[自动化与计算机技术—检测技术与自动化装置]