全谱段成像仪的地表场景模拟

发布时间：2018-10-04 23:19

【摘要】：图像模拟,是在分析和理解传感器成像过程的基础上,通过数学模型来模拟图像获取的过程,从而生成模拟的遥感图像。图像模拟理论现已被应用于多个遥感领域,其中较为重要的用途之一是为传感器的研制及在轨测试提供重要的参考依据,从而保证传感器设计和运行的可靠性。随着图像模拟技术的发展,在可见光或者热红外等窄谱段范围内,该技术得以较快发展,取得了较多的研究成果。但对于全谱段而言,由于其范围较大,且机理复杂,因此模拟难度较大,其图像模拟的研究工作也相对缺乏。针对上述问题,本文开展基于全谱段的地表场景模拟研究。本文主要针对高分四号(GF-4)卫星的在轨测试以及高分五号(GF-5)卫星的研制阶段等问题,利用邻近通道之间的相关性进行图像模拟,从而为GF-4的在轨测试及GF-5的研制提供较为可靠的参考依据。GF-4卫星于2015年12月29 日在我国西昌卫星发射中心成功发射,是中国第一颗地球同步轨道的遥感卫星。为了保证GF-4卫星在轨运行的正常,本文基于HJ星的数据对GF-4卫星进行在轨期间的可见近红外通道地表场景的模拟工作。高分五号(GF-5)卫星按照计划将于2017年发射,该卫星配备高光谱相机、多部大气环境和成分探测的设备。因此,本文基于Aster的数据对GF-5卫星进行全谱段地表场景的模拟工作,从而为发射前的研制阶段提供重要依据。本文所研究的波段范围为全谱段,包括可见近红外波段、短波红外波段、中红外波段及热红外波段,在对现有的地表场景模拟算法研究的基础上,本文提出一种基于邻近通道光谱相关性的地表场景模拟方法。该方法首先考虑数据源通道和待模拟通道间光谱的相关性,在地物光谱信息的基础上,利用传感器的光谱响应函数对光谱信息进行重采样,并采用偏最小二乘线性回归的方法,构建了数据源通道与待模拟通道之间的转换模型。随后使用十折交叉的方法对所构建的模型方程进行误差检验,将误差最小的模型方程作为最终的通道间的转换模型。将模型应用于数据源通道的图像,模拟得到待模拟通道的图像。利用本文所提出的方法,分别对GF-4和GF-5进行地表场景模拟的实验,分别模拟得到相应通道的图像。并运用结构相似度方法对模拟图像和真实图像间的相似程度进行了对比,对模拟结果进行了精度分析,以验证算法的有效性,结果表明,对GF-4卫星可见近红外通道的模拟,其平均相对误差小于5%,相似度在69%～85%之间;对GF-5卫星全谱段通道的模拟,其平均相对误差小于10%。
[Abstract]:Image simulation is based on the analysis and understanding of sensor imaging process, through the mathematical model to simulate the process of image acquisition, so as to generate simulated remote sensing images. Image simulation theory has been applied to many remote sensing fields. One of the more important uses is to provide important reference for the development of sensors and the on-orbit test, so as to ensure the reliability of sensor design and operation. With the development of image simulation technology, in the narrow spectrum such as visible light or thermal infrared, the technology has been developed rapidly, and more research results have been obtained. However, for the whole spectrum, because of its large range and complex mechanism, it is difficult to simulate, and the research work of image simulation is relatively lacking. In order to solve the above problems, the surface scene simulation based on full spectrum is carried out in this paper. In this paper, the correlation between adjacent channels is used to simulate the image of GF-4 satellite in orbit test and the development stage of GF-5 satellite. Thus, the GF-4 satellite was successfully launched at Xichang Satellite launch Center on December 29, 2015. It is the first remote sensing satellite in geosynchronous orbit of China. In order to ensure the normal operation of the GF-4 satellite in orbit, this paper simulates the surface scene of the visible near infrared channel during the orbit of the GF-4 satellite based on the data of the HJ satellite. The GF-5 satellite, which is scheduled to launch in 2017, is equipped with a hyperspectral camera and multiple atmospheric environment and component detection equipment. Therefore, based on the data of Aster, this paper simulates the surface scene of the whole spectrum of GF-5 satellite, which provides an important basis for the research and development of the satellite before launch. The band range studied in this paper is full spectrum, including visible near infrared band, short wave infrared band, middle infrared band and thermal infrared band. In this paper, a surface scene simulation method based on spectral correlation of adjacent channels is proposed. Firstly, the spectral correlation between the data source channel and the channel to be simulated is considered. Based on the spectral information of the ground object, the spectral response function of the sensor is used to resample the spectral information, and the partial least square linear regression method is used. The conversion model between the data source channel and the channel to be simulated is constructed. Then the error test of the constructed model equation is carried out by using the method of ten fold crossing, and the model equation with the minimum error is regarded as the final transition model between channels. The model is applied to the image of the data source channel, and the image of the channel to be simulated is simulated. Using the method proposed in this paper, the surface scene simulation experiments are carried out on GF-4 and GF-5, respectively, and the corresponding channel images are obtained respectively. The structural similarity method is used to compare the similarity between the simulated images and the real images, and the accuracy of the simulation results is analyzed to verify the effectiveness of the algorithm. The results show that the simulation of the visible and near infrared channels of the GF-4 satellite is carried out. The average relative error is less than 5 and the similarity is between 69% and 85%, and the average relative error is less than 10% for the whole spectrum channel of GF-5 satellite.
【学位授予单位】：西安科技大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：P237

【参考文献】