Состоялось очередное заседание общемосковского научного семинара "МАТЕМАТИЧЕСКИЕ МЕТОДЫ АНАЛИЗА РЕШЕНИЙ В ЭКОНОМИКЕ, БИЗНЕСЕ И ПОЛИТИКЕ"
Автор доклада: Герхард Вильгельм Вебер (Middle East Technical University)
Тема: Nonparametric Regression Splines for Regional Atmospheric Correction
16 октября 2013 года в рамках очередного заседания общемосковского научного семинара "МАТЕМАТИЧЕСКИЕ МЕТОДЫ АНАЛИЗА РЕШЕНИЙ В ЭКОНОМИКЕ, БИЗНЕСЕ И ПОЛИТИКЕ" был заслушан доклад на тему "Nonparametric Regression Splines for Regional Atmospheric Correction".
Автор доклада: Герхард Вильгельм Вебер (Middle East Technical University)
In order to obtain high quality data when examining the Earth's surface from a space-based remote sensing platform, correction of land surface reflectance data for the perturbations introduced by the passage of radiation through the Earth’s atmosphere is an important topic within remote sensing. The major methodology in atmospheric correction is to numerically model the whole process of atmospheric attenuation by using a detailed radiative transfer model. However, the high accuracy available with radiative transfer models cannot actually be reached when working on large areas and long time periods, due to unknown atmospheric parameters. In this study, a different approach is represented for the regional atmospheric correction of satellite images by employing nonparametric regression splines within the frame of inverse problems and modern techniques of continuous optimization. Multivariate Adaptive Regression Splines is a well-known nonparametric regression tool from data mining and estimation theory and it is highly capable of handling the nonlinearities and interactions in complex data structures. Conic Multivariate Adaptive Regression Splines, in which the complexity of Multivariate Adaptive Regression Splines is penalized in the form of Tikhonov Regularization and studied as a Conic Quadratic Programming problem, is developed as an alternative method, and it is powerful in overcoming complex and heterogeneous data. It is more model-based and employs continuous, well-structured convex optimization, which uses Interior Point Methods and their codes. In this study, atmospheric correction models obtained through both Multivariate Adaptive Regression Splines and Conic Multivariate Adaptive Regression Splines are applied on a set of satellite images in order to convert the top of atmospheric reflectance values into surface reflectance values. The results are compared with the ones obtained by a radiative transfer-based method.
Дополнительные материалы:
Лекция
Барселона 2014
Глазго 2015