Yield estimation of sugarcane (Saccharum officinarum) from photogrammetry with unmanned aerial vehicles (UAV)

Abstract

Introduction. The estimation of yield variables in sugarcane (Saccharum officinarum) previous to its harvest is important to adapt management practices and to increase the competitiveness of the Costa Rican sugar agroindustry. Objective. The aim of this study was to carry out an estimation of different yield variables in sugarcane through the integration of aerial photos, photogrammetric methods, and calibration of the processed image outputs against measured field data. Materials and methods. A plot with a 5-years old ratoon of the variety B76-256 of sugar cane from CoopeCañita R.L. (Turrialba, Costa Rica) was used for the calibration. The spatial and temporal evolution of crop growth was monitored from May to November 2017 using an Unmanned Aerial Vehicle (UAV) that carried a 16.2-megapixel CoolPix A camera during 6 flights. The photogrammetric processing of the images and the generation of Digital Surface Models (DSM) allowed the estimation of plant height and crop volume. The estimations were correlated with different yield variables as metric tons of fresh biomass per hectare (TMH), kilograms of sugar per ton of fresh biomass (RKA), total kilograms of sugar per hectare (TKA) total kilograms of molasses per hectare (TKM) and yield of kilograms of molasses per ton of fresh biomass (RKM) of the 2017-2018 harvest. Results. The relationship between field-measured yield and the estimated values of plant height was R²= 0.83. The absolute error for yield estimation using the photogrammetric obtained plant height was 3.7 ton.ha^-1, its residual error was about 6.3 ton.ha^-1 and the RMSE was 5.63 ton.ha^-1. Conclusion. According to the results, the best time to predict the sugarcane yield (TMH) through UAV images and photogrammetric tools is around three months before the harvest. No relationship was found between photogrammetric estimations and other sugar content variables of sugarcane.