基于机载高光谱遥感数据的溢油信息提取方法  

Extraction method of oil spill information using airborne hyper-spectral remote sensing data

作　　者：刘丙新[1] 张志达[1] 李颖[1] 陈澎[1] 

机构地区：[1]大连海事大学航海学院,辽宁大连116026

出　　处：《大连海事大学学报》2014年第1期89-92,共4页Journal of Dalian Maritime University

基　　金：国家自然科学基金资助项目(41071260);中国石油天然气股份有限公司科学研究与技术开发资助项目(2011A-0209-01);中央高校基本科研业务费专项资金资助(3132014023)

摘　　要：为减少高光谱遥感数据处理波段,提高处理效率,提出一种基于最小噪声分离波谱的决策树分类方法.利用最小噪声分离变换(MNF)降低数据冗余度,并将图像噪声分离,通过分析各地物的MNF特征值,建立分类决策树,并提取了油膜相对厚度等信息.结果表明,该方法保证了识别精度,有效利用了光谱维信息,明显减小了数据处理时间,从而将高光谱数据用于溢油应急快速产品的生产中.Decision tree classification method was proposed on basis of the minimum noise fraction（MNF） to reduce dimensions of hyper-spectral remote sensing data and improve processing efficiency.The data redundancy was reduced by means of MNF,and the figure noise was separated.The decision tree was established according to analyzing landmarks＇ MNF eigenvalue,and the relative thickness of the oil film was extracted.The results show that the method mentioned could ensure recognition accuracy,achieve effective use of spectral dimension information,as well as reduce the processing time significantly,so as to make the quick products for oil spill response by using hyper-spectral data.

关 键 词：高光谱遥感 溢油监测 最小噪声分离 

分 类 号：TP751.1[自动化与计算机技术—检测技术与自动化装置]