基于光谱曲线形态的高光谱影像检索方法研究

Spectral Curve Shape Feature-Based Hyperspectral Remote Sensing Image Retrieval

随着传感器技术、数据通讯技术的飞速发展,利用各种机载和星载传感器,己经获取到各种不同的海量遥感影像数据。巨大的数据量带来了数据存储和管理的问题,如何实现从海量影像数据中检索出我们所需要的信息显得十分迫切。影像检索最早由Chang于1980年提出,是对传统信息检索的扩展。针对海量遥感影像高效检索的需求和高光谱遥感影像波段数目多的特点,分析了影像检索中的影像距离函数和相似性度量问题,基于经典的曲线简化Douglas-Peucke算法(简称DP算法)提取光谱曲线的形态特征,利用"提取特征"的思想,提出了基于DP算法的光谱曲线和影像检索(简称DPSR)方法,将光谱形态特征应用于影像检索当中。DPSR利用光谱曲线上的特征点,减小了计算量,实现了有效地匹配和检索,适合高光谱遥感影像的光谱检索。文章选择了OMISI高光谱数据的四种易混分地类进行了相似性度量的对比实验。通过与常规的分析方法光谱角匹配(SAM)、光谱信息散度(SID)的对比可以看到,DPSR在较少计算量的情况下能保持较高的计算精度,提供了一种新的影像光谱高效检索方法。此外,文章还提出了尚待进一步研究的问题。

With the rapid development of technology of sensors and data transmission,using all kinds of airplane sensors and satellite sensors,the authors can get different voluminous remote sensing image data of earth.Those voluminous remote sensing image data bring problems of data storage and management.It is becoming increasingly necessary to retrieve some information the authors need from those voluminous image data.Image retrieval was proposed by CHANG firstly in 1980 and can be regarded as expansion of traditional information retrieval.Oriented to the demands of efficient retrieval for voluminous remote sensing image,and considering that there are many bands in hyperspectral remote sensing image,the authors first analyzed image distance function and similarity measure in image retrieval.The most crucial issues in retrieval are spectral features extraction and similarity measure.In the present paper,the authors used classical Douglas-Peucker algorithm（hereinafter referred to DP algorithm） for curve simplification to extract shape features of spectral curve,in order to speed up hyperspectral remote sensing image retrieval.And the authors proposed a new method of spectral curve and remote sensing image retrieval,called Douglas-Peucker Spectral Retrieval algorithm（hereinafter referred to DPSR algorithm）.Spectral shape features were used in image retrieval.DPSR used features of spectral curve,reduced the computation amount,realized match and retrieval efficiently,and is suitable for spectral curve retrieval in hyperspectral remote sensing image.The authors selected four ground features（grass,apple garden,grape garden and pond） in OMISI hyperspectral remote sensing image to compute similarity measure results,in order to test the effect of DPSR algorithm.Compared with traditional analysis method such as spectral angle match（SAM） and spectral information divergence（SID）,DPSR can maintain high precision of results with less amount of computation,and then a new efficient image spectral retrieval method is provided.Besides some additional work the authors need to do in the future,for example,the relationship between threshold,retrieval precision rate and retrieval speed.