Removal of clouds, dust and shadow pixels from hyperspectral imagery using a non-separable and stationary spatio-temporal covariance model

Yoseline Angel; Rasmus Houborg; Matthew F. McCabe

doi:10.1117/12.2241518

25 October 2016 Removal of clouds, dust and shadow pixels from hyperspectral imagery using a non-separable and stationary spatio-temporal covariance model

Yoseline Angel, Rasmus Houborg, Matthew F. McCabe

Author Affiliations +

Proceedings Volume 9998, Remote Sensing for Agriculture, Ecosystems, and Hydrology XVIII; 999824 (2016) https://doi.org/10.1117/12.2241518
Event: SPIE Remote Sensing, 2016, Edinburgh, United Kingdom

Abstract

Hyperspectral remote sensing images are usually affected by atmospheric conditions such as clouds and their shadows, which represents a contamination of reflectance data and complicates the extraction of biophysical variables to monitor phenological cycles of crops. This paper explores a cloud removal approach based on reflectance prediction using multitemporal data and spatio-temporal statistical models. In particular, a covariance model that captures the behavior of spatial and temporal components in data simultaneously (i.e. non-separable) is considered. Eight weekly images collected from the Hyperion hyper-spectrometer instrument over an agricultural region of Saudi Arabia were used to reconstruct a scene with the presence of cloudy affected pixels over a center-pivot crop. A subset of reflectance values of cloud-free pixels from 50 bands in the spectral range from 426.82 to 884.7 nm at each date, were used as input to fit a parametric family of non-separable and stationary spatio-temporal covariance functions. Applying simple kriging as an interpolator, cloud affected pixels were replaced by cloud-free predicted values per band, obtaining their respective predicted spectral profiles at the same time. An exercise of reconstructing simulated cloudy pixels in a different swath was conducted to assess the model accuracy, achieving root mean square error (RMSE) values per band less than or equal to 3%. The spatial coherence of the results was also checked through absolute error distribution maps demonstrating their consistency.

Citation Download Citation

Yoseline Angel, Rasmus Houborg, and Matthew F. McCabe "Removal of clouds, dust and shadow pixels from hyperspectral imagery using a non-separable and stationary spatio-temporal covariance model", Proc. SPIE 9998, Remote Sensing for Agriculture, Ecosystems, and Hydrology XVIII, 999824 (25 October 2016); https://doi.org/10.1117/12.2241518

ACCESS THE FULL ARTICLE

INSTITUTIONAL
Select your institution to access the SPIE Digital Library.

SELECT YOUR INSTITUTION

PERSONAL
Sign in with your SPIE account to access your personal subscriptions or to use specific features such as save to my library, sign up for alerts, save searches, etc.

PERSONAL SIGN IN

No SPIE Account? Create one

PURCHASE THIS CONTENT

SUBSCRIBE TO DIGITAL LIBRARY

50 downloads per 1-year subscription

Members: $195

Non-members: $335 ADD TO CART

25 downloads per 1 - year subscription

Members: $145

Non-members: $250 ADD TO CART

PURCHASE SINGLE ARTICLE

Includes PDF, HTML & Video, when available