基于物理模型的山地光学遥感影像大气、地形和BRDF协同校正
文献类型：学位论文
| 作者 | 王庆芳
|
| 学位类别 | 硕士 |
| 答辩日期 | 2015 |
| 授予单位 | 中国科学院大学 |
| 授予地点 | 北京 |
| 导师 | 李爱农 |
| 关键词 | 山区 Landsat-5 TM 大气校正 地形校正 BRDF校正 HJ-1 CCD |
| 其他题名 | Physics-based atmospheric,topographic and BRDF correction for optical satellite images over mountainous terrain |
| 学位专业 | 地图学与地理信息系统 |
| 中文摘要 | 我国是山地大国，山区面积约占陆地国土面积的70%。利用遥感监测山区的变化与发展具有不可替代的优势，但同时也遇到了巨大的困难与挑战。大气、地形和BRDF效应导致山区光学遥感影像上存在一定程度的辐射畸变，这些辐射畸变在很大程度上影响了影像在地表生物物理参量反演、时间序列数据分析与变化检测等领域的定量应用精度。因此，以消除或抑制影像大气、地形和BRDF效应为目的的辐射校正是山区遥感定量化应用之前必须要做的预处理工作，中高空间分辨率光学影像的辐射校正研究具有重要的实际意义和应用价值。本研究主要探索基于物理模型的山区中高空间分辨率光学影像的大气、地形和BRDF协同校正。 论文主要包括三方面研究内容：（1）选取两种具有代表性的辐射校正物理模型，即基于地表朗伯体假设的Sandmeier模型和考虑地表各向异性散射特性的Li模型，定性、定量地对比分析两种物理模型对于复杂山区、不同太阳高度角Landsat-5 TM影像的地形校正效果，并在此基础上改进Li模型，提高模型的地形校正效果和稳定性。（2）以典型山区的国产HJ-1 CCD影像为研究对象，基于改进模型，对HJ-1 CCD影像进行大气、地形和BRDF的协同校正，并从视觉效果和统计特征方面评估影像辐射校正效果。（3）改进模型依赖MODIS BRDF模型参数产品来解释地表BRDF效应，针对“Landsat-5 TM、HJ-1 CCD影像等与MODIS影像分辨率不匹配影响辐射校正结果”的问题，探索基于HJ-1 CCD影像自身的多期次、多角度观测数据拟合30米尺度的地表BRDF模型数据，并通过由30米BRDF模型数据衍生的HJ-1 CCD NBAR来验证BRDF模型拟合的效果。 论文主要得到以下几点结论：（1）Sandmeier模型在一定程度上抑制了地形效应，但在局地入射角较大的阴坡，由于朗伯假设较大程度偏离地表各向异性散射的真实情况，其过校正现象特别明显；Li模型在地形校正时考虑了地表的BRDF效应，有效抑制了阴坡过校正现象，阴阳坡反射率趋于均一。经改进模型校正的两景影像，近红外波段反射率与光照系数的相关性得到很好的消除；光照区域各波段的类内变异系数减小；改进模型在Li模型的基础上，进一步提高了地形校正效果。（2）经改进模型校正的HJ-1 CCD影像，其整体对比度增强，阴阳坡同类地物反射率的差异得到消除，地形起伏造成的立体感得到有效抑制。可见光波段的反射率分布整体向低值区域移动，蓝光波段移动幅度最大，指示改进模型较好地消除了影像的大气效应；各波段直方图由校正前的双峰分布转变为近似的正态分布，与自然现象中的地物随机性一致。经过改进模型校正后，阴阳坡同类地物的反射率光谱曲线基本一致，改进模型适用于山区环境星影像的辐射校正。（3）基于16天内积累的多角度环境星影像拟合半经验核驱动BRDF模型，反演模型的驱动核权重系数，并重构与环境星影像同尺度的BRDF模型，具有可行性。 本研究为山区中高空间分辨率光学遥感影像的辐射校正自动化处理奠定了基础，为其定量化应用提供了可能。研究的创新点在于，(1)充分考虑了山区地形起伏、地类复杂的特征，首次利用基于NDVI分组的地物的MODIS BRDF模型参数解释中高空间分辨率遥感影像的BRDF效应； (2)首次尝试基于中高空间分辨率的复杂山区环境星多角度观测数据，拟合30米分辨率的BRDF模型。 |
| 英文摘要 | Mountain areas account for approximately 70% of China’s land area, and it is essentially important to monitor the mountain ecosystem change and protect the mountain environment. The development of remote sensing technique has unprecedentedly facilitated the detection of change and development in mountainous areas. Nevertheless, many difficulties and challenges are encountered in the remote sensing applications in mountainous areas. Atmospheric, topographic and BRDF effects lead to radiometric distortions in optical remote sensing images of mountainous areas, and these distortions severely impact the accuracy of images’ quantitative applications, like biophysical parameter inversion, time series data analysis and change detection. Therefore, radiometric correction, aiming to eliminate or at least suppress the atmospheric, topographic and BRDF effects of mountain-area images, is a preliminary and necessary step for the subsequent quantitative applications. Radiometric correction research holds great practical significance, and this paper mainly focuses on combined atmospheric, topographic and BRDF correction for moderate and high resolution satellite imagery of mountain areas using physics-based model. The content of this paper mainly contains: (1) physics-based topographic correction model comparison and modification. Sandmeier model (representative of model with Lambertian assumption) and Li model (representative of model accounting for the anisotropic behaviors of land surface scattering) are selected to test and evaluate the topographic correction effects with images of typical mountain areas acquired under different sun elevation angles. Moreover, Li model is modified and the modified model performs very well and stably. (2) HJ-1 CCD images’ radiometric correction. Based on modified model, combined atmospheric, topographic and BRDF corrections are implemented for a typical mountain-area HJ-1 CCD scene. The radiometric correction effects are evaluated visually and statistically. (3) BRDF model inversion based HJ-1 CCD images. The modified model relies on MODIS BRDF parameters to account for land surface BRDF effect, and the resolution difference between Landsat-5 TM-like image and MODIS image potentially impacts model’s performance. Based on multi-date and multi-angular HJ-1 CCD images, 30-m resolution BRDF model is constructed, and the derived HJ-1 CCD NBAR data are employed to validate its effect. The main conclusions are summarized as : (1) Sandmeier model suppresses topographic effect to some extent. However, serious overcorrection occurs when local incidence angles are relatively large, which can be attributed to deviations from reality under Lambertian assumption. Li model reduces the overcorrection effectively for the sun-averted slopes and achieved satisfactory results. The modified model successfully eliminates the correlation between Near Infrared band reflectance and illumination condition, and reduces the coefficients of variance of illuminated forest reflectance for each band. The modified model further improve the topographic correction effect. (2) The modified model performs well in the radiometric correction of HJ-1 CCD imagery of mountainous areas. After correction, the overall contrast of the image is enhanced; reflectance of the similar land cover type on sun-facing and sun-averted slopes tends to be consistent and terrain relief is effectively restrained. The overall reflectance distributions for visible bands move towards lower values, and this is especially true for blue band, which indicates that atmospheric effect is successfully removed. Reflectance histograms for each band turn from bimodal distribution to normal distribution, which is accordant with the randomness of land targets. (3) It is feasible to fit BRDF model based on multi-date and multi-angular HJ-1 CCD observations. And through this, the 30-m resolution BRDF model matching HJ-1 CCD and Landsat-5 TM images can be obtained. |
| 语种 | 中文 |
| 源URL | [http://ir.imde.ac.cn/handle/131551/13960]  |
| 专题 | 成都山地灾害与环境研究所_数字山地与遥感应用中心 |
| 作者单位 | 中国科学院成都山地灾害与环境研究所 |
| 推荐引用方式 GB/T 7714 | 王庆芳. 基于物理模型的山地光学遥感影像大气、地形和BRDF协同校正[D]. 北京. 中国科学院大学. 2015. |
入库方式： OAI收割
来源：成都山地灾害与环境研究所
浏览0
下载0
收藏0
其他版本
除非特别说明，本系统中所有内容都受版权保护，并保留所有权利。