作　　者： ; ; ; ;

机构地区： 北京航空航天大学仪器科学与光电工程学院

出　　处： 《科技致富向导》 2013年第15期67-68,108,共3页

摘　　要： 高光谱遥感影像具有光谱分辨率高、图谱合一的特点，影像中的地物组成也多种多样，针对图像数据中目标出现概率较小、检测概率较低等问题．本文提出了基于超平面构建的高光谱数据目标异常检测方法。它通过对高光谱数据的结构分析，构建出高维空间中的背景超平面算子．以待测像元在其正交子空间的投影值来表征异常程度的大小。通过OMIS和AVIRIS实测数据验证，并与RX算法相比较，基于超平面构建的高光谱异常检测算法能够在多种背景环境下检测出纯度高、光谱差异大的异常像元，并具有较高的检测率和较低的虚警率，因此具备了有效性和较好的实用性。’ Hyperspectral remote sensing spectrometer can provide images with high spectral resolution by collecting many spectral continuous images at the same time. It comes up to the hyperspectral image anomaly detection algorithm based on hyperplane structure ,which can well solve the targets detection that appear in lower percent, smaller volume, and heavier disturbance in many kinds of surface features. Through the analysis of the Hyperspectral information, it puts forward the background hyperplane operator which does work by projecting the pixel into orthogonal subspace to show the level of anomaly. By the full experiments of OMIS and AVIRIS, compared with RX algorithm, hyperspectral image anomaly detection algorithm based on hyperplane structure dose a good job in complex surface features, Moreover, it works in lower alarm and better detection percentage to be more effective and widely used.

关 键 词： 高光谱 异常检测 超平面

领　　域： [自动化与计算机技术] [自动化与计算机技术]