Extraction of Rice Phenological Metrics Using Temporally Correlated Multispectral Drone Imagery

The rice crop holds great potential to contribute to the economy of most South Asian agricultural countries. However, due to global warming, changes in weather and climate have made it difficult for farmers to accurately predict the timing of various growth stages in the life cycle of rice. Deciding when to apply fertilizers to maximize yield and pesticides to avoid diseases becomes a challenge with conventional farming techniques. Hence, extracting key phenological metrics and analyzing the growth dynamics of different rice varieties are essential for this purpose. The foremost issue in this process is the lack of available multispectral drone imagery datasets, especially in South-East Asia. Therefore, a novel multispectral dataset of rice crops is presented that consists of three different varieties: Kainat, Hybrid, and Super Kernel in East Punjab, Pakistan. The study’s objective is to use this dataset to model a time series of Normalized Difference Vegetation Index (NDVI) and use the change detection method to map key phenological variables using the rate of change of NDVI. The analysis of available data using phenological metrics represents that the Kainat Basmati variant is growing faster than Hybrid as it reaches its peak about ten days earlier than Hybrid. Overall, our study demonstrates that remotely captured unmanned aerial vehicle (UAV) imagery of rice can streamline the process of predicting phenological metrics of rice and provide farmers with an automated statistical model of the growth stages of various crops.