Authenticate
For signed up users

Remote sensing data provide operative and unbiased information about vegetation state and dynamics. Remote sensing data time series analysis facilitates vegetation types discrimination. However, haze, clouds and their shadows distort spectral reflectance values in Visible, Red and NIR bands. Presently, these hindering factors are well dealt with during data pre-processing and multi-temporal image composing, resulting in data gaps appearing within time series, but residual impurities and instrument noises still disarrange data in time series. Current approaches for time series smoothing deal mostly with data disturbances due to noises and hindering factors, while time series are considered gapless. Besides, these methods do not imply detection and following exclusion of certainly disturbed data. However, data gaps and distortions must be considered jointly to avoid drawbacks of data recovery process through distorted data. This paper describes the use of local polynomial approximation within moving window of variable size jointly for time-series smoothing and data gaps filling.