BJ-1智能小卫星多曝光量数据特征及其积雪提取方法研究

摘要

BJ-1 micro satellite can adjust CCD exposure by altering the imaging integration time,which not only can affect the digital numbers(DN) and the image quality but also can bring about the uncertainties in retrieving the land surface parameters quantitatively.Taking the HJ-1 CCD datasets as references,the authors calculated and analyzed the spectral characteristics and image quality of the high reflective targets in BJ-1 CCD images.Some conclusions have been reached:The image quality will be improved with a moderate exposure and impaired with over-exposure.The spectral features of the snow in terrain shadow can be enhanced with the increasing imaging integration time,while the spectral characteristics of the snow on the sun side would be weakened as a result of the image "saturation".On the basis of the spectral features of the snow and the data characteristics of the multi-exposure BJ-1 CCD imageries,a normalized model of the multi-exposure imageries for extracting the snow cover is proposed with the accuracy (classification success index,CSI)of 89.95％,which is superior to the accuracy (CSI)of 82.25％ of mono-exposure model and the accuracy(CSI) of 75.95％ of supervised classification.The results achieved by the authors will greatly facilitate the design of the more target-specific intelligent sensor and the development of more accurate inversion theory and techniques in retrieving the land surface parameters quantitatively.%通过调控成像积分时间改变曝光量,会在改变图像DN值和图像质量的同时,给地面参数定量反演带来很大的不确定性.以环境星(H J-1)CCD数据为参照标准,以积雪为研究对象,系统分析了北京一号小卫星(BJ-1)多曝光量数据的积雪图像质量及其光谱特征变化规律;在模拟B J-1多曝光量数据的基础上,提出了面向多曝光量数据的积雪提取方法,并评估了不同算法的积雪面积提取精度.结果表明,BJ-1 CCD数据的积雪图像质量随曝光量增加有所改善,但过分曝光也会导致图像质量下降;阴影区积雪的光谱差特征随曝光量的增加而增强,向阳面积雪的光谱差特征因“饱和”而大大削弱.基于BJ-1模拟数据,提出了面向多时相、多曝光量数据的归一化积雪提取模型,该模型的分类成功指数(classification success index,CSI)达到89.95％,优于单一曝光量的82.25％和传统监督分类的75.95％的提取精度,为研发更具目标针对性的智能传感器和高精度地表参数遥感反演算法提供了有益的借鉴.