SUGARCANE YIELD PREDICTION USING REMOTE SENSING AND MACHINE LEARNING TECHNIQUES: A SYSTEMATIC REVIEW OF THE LITERATURE.

Abstract

The use of remote sensors and machine learning models have become an important decision-making tool for predicting sugarcane yield. In this work, a systematic review of scientific literature was carried out, extracting and synthesizing different indicators used in sugarcane yield prediction studies. In this review, 386 relevant studies were obtained from five electronic databases, selecting 47 studies for in-depth analysis using exclusion and selection criteria. The selected studies were investigated, analyzing the methods, variables used, and making suggestions for future research. According to the analyses, the most used variables are climatic variables (temperature, precipitation and evapotranspiration), and crop variables such as the number of cuts. The most used algorithms are random forests and multiple linear regression. The most used remote sensors are satellites such as Sentinel 2 and Landsat 8. The most used vegetation indices are NDVI, GNDVI; SAVI and EVI. It is reported that the use of spectral bands such as near infrared and short wave, in some cases help explain the yield of sugar cane.