Application of RGB UAV images to identify spectral patterns and estimate rice production

Abstract

Monitoring rice plant growth is crucial for evaluating rice field management and yield production. RGB images are generated from Unmanned Aerial Vehicles (UAV) with RGB cameras. UAVs produce high spatial and temporal resolution, while RGB cameras are commonly used and cheap. The objectives of this study were to identify the spectral pattern of rice plant growth and to estimate yield production based on the spectral value of RGB images. The spectral pattern and yield estimation were analyzed using confidence interval (CI) and regression, respectively. Results show that spectral pattern during the vegetative until ripening stage forms a concave with minimum value in the generative stage and decreases towards the harvest stage. Based on the CI value, the high interval between upper and lower happened in the vegetative and ripening stages while the low interval happened in the generative stage. The high CI in the vegetative and ripening stages was due to the soil background and complexity of the rice plant canopy, respectively while the low CI in the generative stage was due to the homogeneous response of the leaf canopy. The best rice yield estimation based on the spectral value occurs in the ripening stage with an R² of 0.84.

Keywords: chlorophyll content, confidence interval, drone images, rice plant, regression