多元变化检测的相对辐射校正方法研究

Relative radiometric correction for remote sensing images based on multivariate alteration detection

基于多元变化检测的相对辐射校正方法通过阈值获取校正点,该方法的校正结果优于传统人工参与的校正方法。文章对方法中阈值选择以及自然景观特征等影响因素做了进一步研究,选择北京市平原区和山区的各2期TM影像作为数据源,运用均方根误差和变异系数2个统计特征参数比较和评价校正结果,结果表明:基于多元变化检测的相对辐射校正方法获得的结果有利于后续数据分析;不同阈值获得的校正结果没有明显差异;不同自然景观特征对该方法影响程度不同。

The relative radiometric correction is an important step of processing and analyzing of multi-temporal remote sensing images.The relative radiometric correction based on multivariate alteration detection（MAD） obtained calibration points by threshold,and the result by the method was superior to the traditional correction methods requiring manual manipulation.The paper did further research at the factors affecting the results of radiometric correction from threshold selection and scene characteristics,selected 1997 and 2004 Landsat TM images of plain area and mountain area in Beijing respectively,employed the root mean square error and the coefficient of variation for comparing and evaluating the correction results.The results showed that the result by the correction method based on MAD was beneficial to follow-up data analysis;the correction results of different thresholds were not significantly different;the different scene characteristics had different effect on the correction results.