森林资源监测中 GF-2卫星影像波段配准误差分析

摘要

Based on the GF‐2 satellite images orthophoto correction and band simulation registration error test ,this paper analyzes the selection of GF‐2 satellite orthorectification method and the influence of GF‐2 satellite image automatic classification results under the different registration error .Finally ,it introduces a primary test of GF‐2 remote sensing image in forest resources monitoring applications .The results show that :for ortho correction ,when the correction accuracy is required to be controlled at RMS < 2 ,the number of control points in the range of 85~95 is more reasonable ,and w hen the number of control points is 90 ,the RMS value is the smallest ;after comparing the rational function model and satellite orbital model ,the satellite orbital model has the higher correction accuracy ;based on visual interpretation ,cubic convolution sampling method can output the best results .For band simulation registration error test ,there is a significant linear relationship between the registration error and different land area change ;and for the forest area monitoring ,the registration error shall be less than 0 .3 pixels .The paper tries to provide a reference for the application of new domestic satellite data to forest resources monitoring .%通过对GF‐2卫星影像正射校正及波段模拟配准误差试验，分析GF‐2卫星正射校正方法的选择以及不同配准误差下对GF‐2卫星影像自动分类结果的影响；最后介绍GF‐2遥感影像在森林资源监测应用中的初步测试。研究结果表明：正射校正时，当校正精度要求控制在RM S＜2时，控制点数量选择范围在85～95间较为合理，且控制点数在90个时，RM S值最小；经有理函数模型与卫片模型比较后，卫片模型校正精度较高；以目视判读为主时，实践中建议使用三次卷积重采样法输出结果最好；波段模拟配准误差试验中，配准误差与各地类面积变化间存在显著的线性关系；对于森林面积监测时，配准误差应小于0．3个像元。此研究可为新型国产卫星数据在森林资源监测中的应用提供参考。