基于半经验模型的遥感影像地形校正效应及适用性分析  

Effect and Applicability of Terrain Correction of Remote Sensing Images Based on Semi-Empirical Models

作　　者：晏红波 王佳华 卢献健 李浩 YAN Hongbo;WANG Jiahua;LU Xianjian;LI Hao(College of Geomatics and Geoinformation,Guilin University of Technology,Guilin 541004,Guangxi,China)

机构地区：[1]桂林理工大学测绘地理信息学院,广西桂林541006

出　　处：《山地学报》2023年第5期759-770,共12页Mountain Research

基　　金：国家自然科学基金(42361052);广西自然科学基金(2022GXNSFBA035639);广西空间信息与测绘重点实验室开放基金(桂科能19-050-11-23)。

摘　　要：地形校正是遥感影像数据处理的关键步骤,但校正模型在不同地形条件和不同尺度下的校正效果仍缺乏充分、全面验证和比较,限制了地形校正模型的广泛应用。为了探究不同地形条件下地形校正模型的效果和适用性,本研究选取Landsat 8 OLI 30 m影像数据作为数据源,并结合SRTM_V330 m DEM数据,采用C、SCS+C(Sun-Canopy-Sensor+C)、Minnaert三种半经验地形校正模型在高原、山地、盆地、峰丛四种典型地形进行校正实验;运用目视分析、相关性分析和统计分析法,对上述方法在不同地形样区的校正效果进行了深入分析和对比。结果表明:(1)在高原样区,C模型的校正效果最好,能够显著减小影像反射率与太阳入射角余弦值之间的线性回归斜率和决定系数(R 2),并且各波段的四分位距减少量也最多,尤其在近红外波段,减少了26%。(2)在山地样区中,SCS+C模型考虑了植被生长的向地性,校正后R 2最小,为0.003,且各波段IQRR(Interquartile Range Reduction)也最大,说明其在山地样区适用性更高。(3)在盆地样区中,Minnaert模型考虑了地表的双向反射特性,校正后R 2最小,为0.008,说明其对影像反射率与太阳入射角余弦值之间相关性的削弱程度最大,IQRR也比其他两种方法更大,因此其对盆地样区具有较好的适用性。(4)在峰丛样区中,C和SCS+C模型校正后出现了过校正现象;相比之下,Minnaert模型的校正效果较好,但仍需进一步改进以提高对阴影区域的校正效果。本研究可为遥感影像数据处理提供技术支撑。Terrain correction is a necessarystep in interpreting remote sensing images.The performances of terrain correction modelsquite depend on varied terrains and image scales,whichrestricted applicability of these models.Unfortunately,there was still a lack of comprehensiveverification and comparison of theseterrain correction models.This study aimed to explore the applicability of terrain correction models for different terrains.Three commonly used semi-empirical terrain correction models,C,SCS+C and Minnaert,were evaluated underfour typical terrains of remote sensing images(plateau,mountain,basin and peak).The correction effect of the models under different terrain conditions were compared by using visual,correlation and statistical analysis methods.This research found(1)the C model had the best performance in plateau area,which was able to reduce the linear regression slope and coefficient of determination(R 2)between image reflectance and the cosine of solar incidence angle significantly.It highlyreduced the interquartile range(IQR)of each band,especially in the near-infrared band,with a decrease by 26%.(2)The SCS+C model was more applicable in mountainous areas because it accounted for the geotropic nature of trees(vertical growth).It had a low coefficient of determination of 0.003,with a large reduction in the interquartile range for each band.(3)The Minnaert model took into account of the bidirectional reflectance of surfaces,which improved its correction effect in basin sample areas.After correction,it had a low correlation value of 0.008,which indicated that this model weakened the correlation between image reflectance and the cosine of solar incidence angle more than the other two models.It also had a larger IQR reduction than other models.(4)In peak cluster area,the images were over correction by the C model and the SCS+C model;the Minnaert model had better correction performance but further improvements were needed in shadow region.This study provides technical support for data processing of remote sensing

关 键 词：地形校正 遥感影像 半经验模型 适用性 

分 类 号：P23[天文地球—摄影测量与遥感]