归一化转动惯量特征的遥感影像检索  

Remote sensing image retrieval based on normalized moment inertia characteristic

作　　者：魏立飞[1] 刘文凤[2] 张帆[1] 林丽群[1] 

机构地区：[1]湖北大学资源环境学院,武汉430062 [2]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》2017年第2期115-119,134,共6页Science of Surveying and Mapping

基　　金：长江科学院开放研究基金项目(CKWV2012325/KY);国家自然科学基金项目(61201341;41371344);干旱气象科学研究基金项目(IAM201512);农业部农业信息技术重点实验室开放基金项目(2013004);数字制图与国土信息应用工程国家测绘地理信息局重点实验室开放研究基金项目(GCWD201407)

摘　　要：针对传统遥感影像检索方法大多存在精度不高、效率低下等不足的问题,该文提出了一种基于归一化转动惯性特征的遥感影像检索算法。该算法对经过脉冲耦合神经网络处理的二值影像序列进行计算,提取影像序列的归一化转动惯量特征;同时,利用马氏距离结合Pearson积矩相关法来度量各特征矢量之间的相似性,提高检索结果的正确率。实验结果证明,该算法有效地兼顾了影像的内容结构信息,不仅可以快速地进行检索计算,还能提高检索精度。Aiming at the low accuracy and poor efficiency of traditional remote sensing image retrieval methods, this paper proposed a remote sensing image retrieval method based on normalized moment inertia characteristics. The binary image sequence, which was obtained by pulse-coupled neural networks, was calculated to extract normalized moment inertia characteristics; meanwhile, the Mahalanobis distance and the Pearson product-moment correlation method were combined to measure the similarity between the feature vectors to improve the accuracy of retrieval results. Experimental results showed that this method effectively considered the image structural informa- tion, which could quickly and efficiently provide accurate retrieval results.

关 键 词：影像检索 脉冲耦合神经网络 归一化转动惯量特征 

分 类 号：TP751[自动化与计算机技术—检测技术与自动化装置]