遥感图像大气校正光谱矢量分析模型

The Spectrum Vector Analysis Model(SVAM ) put forward by this project is fit for remote sensing images atmospheric correction ranging from visible light to near infrared. Structuring based on the principle of transmit of light, this model meanwhile took terrain factor, BRDF factor of objects and rating factor of every type of objects in mixed pixel into consideration, and eliminated all these three factors during the model solution . The corrected image free from these three factors can adapt to terrain change, BRDF feature of objects and the variety of mix rate in mixed pixels. In corrected images, mountain areas equal to flat, mixed pixels equal to pure ones and Non Lambert pixels equal to Lambert ones, so the effects of the image keep the same as a whole. This model improved the method for atmosphere parameter acquiring, improved the method for correcting mutual radiation of near pixels, improved the method for stripping atmosphere image, invented a new method for atmospheric correction which cost is relatively low, adaptation is relatively good. This Model is widely needed by customs so that it has a great market. Besides, physical meanings of corrected images are clear, atmospheric correction precision of correction is hopefully better than the one generated by the most excellent model with the same constrained conditions. As a result, this model is applicable largely. SVAM demonstrated features of spectrum of objects with a series definition, theorem and deduction wich is the base of theories of atmospheric correction, as well as classifications, recognition and matching of spectrum of objects. So it has theoretical values too.

本项目创建的光谱矢量分析模型适用于可见光到近红外波段卫星遥感图像大气校正，模型根据光的传播机理构建，同时考虑了地形因子、地物BRDF因子、混合像元中每种地物占比因子，模型求解中消除了这三个因子，无这三个因子数据情况下校正图像大气影响能适应地形变化、能适应地物BRDF特征、能适应混合像元混合比变化，校正的图像山区与平地等效、混合像元与纯像元等效、非朗伯体像元与朗伯体像元等效、全景效果一致。模型改进了大气参数获取方法、改进了邻近像元交叉辐射校正方法、改进了大气剥离方法、创新了大气校正方法，大气校正成本较低、适应性较好、实用性较好、用户需求迫切、市场较大，大气校正结果图像物理含义明确,校正精度可望超过同约束条件最优秀大气校正模型，有较大应用价值。 项目中论证了地物光谱的性质，其中的一系列定义、定理、推论和结论是大气校正的理论基础，也是地物光谱分类、识别、匹配的理论基础，有理论价值。

大气对阳光的散射、吸收和反射，导致卫星对地观测遥感图像中地物模糊不清，本项目采用光谱分析模型（GPFX）将遥感图像分离成地物图像和大气图像，对地物图像进行大气校正，获得了清晰的地物图像。GPFX大气校正精度、效率、成本及适用范围均有明显比较优势。.GPFX完全根据图像自身光谱信息进行大气校正，不必要数字地形模型，不必要同步大气测量参数及地物测量参数，能适应地形变化及地物BRDF特征因子变化，校正后的图像精度明显优于目前流行的大气校正模型校正的图像精度,地形起伏区与平坦区精度相当,全景图像色彩均衡，亮度均衡。GPFX校正的地物图像及输出的大气图像像元值均为反射率，均可合成真彩色图像，均可用于定量分析，应用范围广泛。GPFX校正的地物图像中保留了原图像的地形信息及BRDF信息，因此，图像经过辐射定标后，宜用GPFX进行大气校正。GPFX只要求原图像中有（红，蓝）两个波段图像就可对全部波段图像进行大气校正，因此可用于大部分遥感图像大气校正。GPFX实验程序在i5 3230M CPU中执行基本能满足科研工作效率,表明大气校正工作效率有明显比较优势、根据GPFX研发的卫星遥感图像大气校正软件有明显比较优势。