SEPARABILITY OF SUGARCANE DYNAMIC USING BASIC STATISTICS METRICS FROM SATELLITE TIME-SERIES

In this study, we have done a comparison among different
machine learning algorithms to develop a model to identify
ratoon, expansion, under renovation, and renovated
sugarcane areas using basic statistics metrics from the
oli/landsat-8 time-series data, along with the data from srtm.
this study was in a region among São Paulo and Minas Gerais
states for the crop-season 2020/21. we have used three
datasets, one balanced and the others not. our methodology
has the potential to separate the proposal sugarcane classes.
also, feature selection is an important step to decrease the
computational cost and improve the model accuracies. the
approach to generate simple statistical metrics based on time
windows is useful because we can reduce the size of our data
and help us summarize all the variability of the interesting
areas in these different periods.