Monitoring of Rice Growth with UAV-derived Aerial Imagery

doi: 10.14456/mijet.2019.4

Abstract

This study aims to monitor rice growth by using the reflectance relation of (R-B)/(R+B) and R/(R+G+B) to predict rice biomass before and after the heading stage. UAV-derived aerial imagery was obtained from an RGB camera attached on the UAV, which flew to take pictures at the altitude of 90 meters, with the front-overlap of 90% and side-overlap of 60%, to be used for calculating Green-Red Vegetation Index (GRVI) and Red Green Blue Index (RGBI). In addition, the field data were divided into two parts data for calibration and data for evaluation of three models through Rapid minder Studio 9.1, namely Generalized Linear Model, Deep Learning, and Random Forest. 120 sets of field biomass data were collected, 80 of which were for calibrating the models, and 40 for evaluating the models. After the evaluation of Coefficient of Determination (R²) and Root Mean Square Error (RMSE), for the biomass of rice before the heading stage, it was found that for GRVI, R² and RMSE were 0.920 and 0961, respectively, and for RGBI, R2 and RMSE were 0.918 and 0.697, respectively. Meanwhile, for the biomass of rice after the heading stage, it was found that for GRVI, R² and RMSE were 0.854 and 1.648, respectively, and for RGBI, R² and RMSE were 0.810 and 1.530, respectively. For both periods, the most suitable prediction model was Random Forest. This shows that the reflectance relation of both equations based on GRVI and RGBI could be used to monitor rice growth.