|
IMAGE PROCESSING, PATTERN RECOGNITION
Atmospheric correction of hyperspectral images using small volume of the verified data
A. Yu. Denisovaab, V. V. Myasnikovab a Samara National Research University, Samara, Russia
b Image Processing Systems Institute îf RAS – Branch of the FSRC “Crystallography and Photonics” RAS, Samara, Russia
Abstract:
In this article, we propose a novel method for atmospheric correction of hyperspectral imagery. At the first stage, the atmospheric correction parameters are derived from a scene image using a well-known radiation transfer model. In contrast to the other methods, we apply the standard equation of radiation transfer in a nonlinear form to describe atmospheric effects and a linear spectral mixture model to describe the unknown undistorted hyperspectral image. Applying both of these mathematical models simultaneously, we estimate the parameters of atmospheric correction using the hyperspectral image itself and the verified data about the registered scene. The verified data is taken to mean a set of (undistorted) spectral signatures, which can be presented in different linear combinations in the registered scene. Neither precedential information (a set of pixels containing predefined spectral signatures) nor pure hyperspectral pixels are required in our method. Therefore, the proposed method can be applied for the identification of a model of atmospheric distortions and their subsequent correction. The experimental results presented in the article demonstrate qualitative characteristics of the proposed method.
Keywords:
Earth remote sensing, radiation transfer equation, hyperspectral images, spectral signatures, spectral profile, linear spectral model.
Received: 19.08.2016 Accepted: 26.08.2016
Citation:
A. Yu. Denisova, V. V. Myasnikov, “Atmospheric correction of hyperspectral images using small volume of the verified data”, Computer Optics, 40:4 (2016), 526–534
Linking options:
https://www.mathnet.ru/eng/co247 https://www.mathnet.ru/eng/co/v40/i4/p526
|
Statistics & downloads: |
Abstract page: | 178 | Full-text PDF : | 60 | References: | 32 |
|