基于掩膜的光谱二阶导数法目标提取  

Second Order Derivatives of Spectrum Based on Mask Algorithm

作　　者：王钦军[1] 蔺启忠[1] 黎明晓[2] 

机构地区：[1]中国科学院遥感应用研究所,北京100101 [2]中国地震局分析预报中心,北京100036

出　　处：《遥感信息》2006年第4期15-19,共5页Remote Sensing Information

基　　金：国家自然科学基金项目"农业环境土壤地球化学的遥感机理研究"(40371085)资助

摘　　要：提出一种简单实用的目标提取方法:基于掩膜的光谱二阶导数法。它是在掩膜的基础上对目标光谱和像元光谱进行二阶导数编码后计算两条光谱曲线的相似度,然后,根据相似度与识别误差来判别像元的归属,从而达到识别目标的目的。试验结果证明,由于综合使用了光谱强度和光谱波形特征,相对光谱二阶导数法而言,最高精度的平均值提高了34%,平均精度的平均值提高了43%,同时,目标提取试验结果也表明它是一种有效的目标提取算法。A new algorithm named Second order Derivatives of Spectrum Based on Mask is presented.There are two steps : one is to encode the target spectrum and the pixel one according to their second order derivatives,another is to calculate the similarity between these two encoded spectra and encode the image according to the similarity and identification error.Results from experiments showed that because it simultaneously used the spectral intensity and waveform characteristics,so its precision is high and average precisions were improved 34% and 43% respectively.At the same time,target extraction experiments showed that it is an effective target detection algorithm.

关 键 词：掩膜 二阶导数 目标提取 

分 类 号：P237[天文地球—摄影测量与遥感]