Prediction of Sugarcane Sucrose Content and Optimal Harvest Date Using Multi-Spectral Time Series Image Processing of Satellite Data

This work presents a method for predicting the sucrose content (pol value) of sugarcane crops using multispectral time series satellite imagery. The methodology utilizes spectral signatures from sugarcane crops to calculate the Normalized Difference Vegetation Index (NDVI). It then integrates this information with environmental parameters such as temperature, precipitation, and agronomic data such as date of planting, age of the cane. Three models were explored i.e. Nonlinear Least Square Regression, Random Forests and Gradient Boosting Machines (GBM). Multispectral images from the Planetscope (3-m resolution) during the season Jan-2020 to March-2022 were studied. Ground truth data for pol values were collected from 216 fields. Results show a negative correlation between pol and both NDVI and temperature. There was a positive relation between pol and age. Highest accuracy obtained through GBM model with R² of 0.94 (test dataset) with MSE (Mean Square Error) 0.44 and R² of 0.74 (validation dataset) with MSE 0.96.