Processing and Use of Satellite Images in Order to Extract Useful Information in Precision Agriculture

Abstract

Image analysis methods were developed and diversified greatly in recent years due to increasing speed and accuracy in providing information regarding land cover and vegetation in urban areas. The aim of this paper is to process satellite images for monitoring agricultural areas. Satellite images used in this study are medium and high resolution images taken from QuickBird and SPOT systems. Based on these images, a supervised classification was performed of a very large area, having as result the land use classes. Supervised classification can be defined as the ability to group the pixels that compose the satellite image, digitally, in accordance with their real significance. Gaussian algorithm of maximum similarity (Maximum likelihood) was used, referred to in the specialty literature as maximum likelihood method or probabilistic classification, and based on the use of probability theory (function Gaussian) to compare the spectral values of each pixel in hand with statistical " fingerprint "of each area of interest. Practically, conditional probabilities were calculated of belonging to one class or another. The points in the middle of the group have a higher probability of belonging to the certain class, probability intervals (concentric isolines or contours of equal probability) being delimited graphically by izocontours expressing spectral variations within each set of training.

Assessment of the Ecological Health of Afrotropical Rivers Using Fish Assemblages: A Case Study of Selected Rivers in the Lake Victoria Basin, Kenya

“…Land use classification and catchment maps were generated with QGIS 3.14, using Semi-Automatic Classification plugin to download sentinel-2 images during the study period (Congedo, 2020a,b). The satellite images were preprocessed and processed for all the band sets, mosaic, clipped and supervised classifications were done for land use/land cover (Akgün et al, 2004;Huth et al, 2012;Congedo, 2016;Herbei et al, 2016) in each of the four catchments, with four categories of land use (forest, grassland, cropland and shrubland). Water quality measurements were explored before further analysis using box-and-whiskers plot to visualize summaries and compare their variation (Williamson et al, 1989;Dekking et al, 2005;Hubert and Vandervieren, 2008) at the catchments.…”