Remote Sensing tools and approaches are now widely used in many different fields such as mining engineering, soil sciences, environmental monitoring, etc. Satellite data is useful within many geosciences because it provides a large amount of time-space continuous data which is easily and quickly accessible. Clouds and shadows within satellite images, however, are an important challenge as they obscure surface features. In many cases, the image cannot be used or the target area cannot be studied because of cloud cover. To counteract this, Geostatistical tools can be used to estimate the missing data in the target area. In this study, a Sentinel-2 image from Copernicus data of land cover in Emilia Romagna (Italy) is used and the spectrum bands are resampled into 10-m resolution. The objective of the study is to estimate the RGB-values of a cloud-covered area within the image. The statistical parameters and the spatial variability of nearby pixels are studied by testing different neighborhoods to analyze the possibility of interpolation within the cloud-covered area. Then the estimator properties are controlled and the values of the cloud-covered pixels are estimated by assuming different parameters such as the size and shape of the neighborhood in consideration, number of pixels used for the estimation and the pixels distribution in the image. The reliability of results is then evaluated by analyzing the estimation variance of each estimated pixel and mapping them. The analysis has been carried out using mainly MATLAB and VBA programming. To validate the results, an image taken at a similar a timeframe has been used for comparison. Results show the advantages and disadvantages of different ways to estimate the cloud-covered area and reliability of the estimations are compared. In addition, by comparing the results with the original values, the effect of the cloud cover has been evaluated on different remote sensed band values.
Using Geostatistical approaches to estimate the RGB values under a cloud covered area within the Sentinel-2 Images
Ashkan Tayebi Gholamzadeh
;Roberto Bruno;Sara Kasmaeeyazdi;Francesco Tinti
2022
Abstract
Remote Sensing tools and approaches are now widely used in many different fields such as mining engineering, soil sciences, environmental monitoring, etc. Satellite data is useful within many geosciences because it provides a large amount of time-space continuous data which is easily and quickly accessible. Clouds and shadows within satellite images, however, are an important challenge as they obscure surface features. In many cases, the image cannot be used or the target area cannot be studied because of cloud cover. To counteract this, Geostatistical tools can be used to estimate the missing data in the target area. In this study, a Sentinel-2 image from Copernicus data of land cover in Emilia Romagna (Italy) is used and the spectrum bands are resampled into 10-m resolution. The objective of the study is to estimate the RGB-values of a cloud-covered area within the image. The statistical parameters and the spatial variability of nearby pixels are studied by testing different neighborhoods to analyze the possibility of interpolation within the cloud-covered area. Then the estimator properties are controlled and the values of the cloud-covered pixels are estimated by assuming different parameters such as the size and shape of the neighborhood in consideration, number of pixels used for the estimation and the pixels distribution in the image. The reliability of results is then evaluated by analyzing the estimation variance of each estimated pixel and mapping them. The analysis has been carried out using mainly MATLAB and VBA programming. To validate the results, an image taken at a similar a timeframe has been used for comparison. Results show the advantages and disadvantages of different ways to estimate the cloud-covered area and reliability of the estimations are compared. In addition, by comparing the results with the original values, the effect of the cloud cover has been evaluated on different remote sensed band values.I documenti in IRIS sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.
