A Deep Learning Approach for Yield Estimation and Phenotype Analysis in Rice Crops

Phenomics is an emerging field of study in India which integrates biology with transdisciplinary techniques including image processing, data science, and engineering. This technology became popular in many countries to scrutinize plants at different growth stages for their photosynthetic efficacy and climate resilience. Over the years, a glut of studies has investigated developing a robust technique to predict the production and yield of rice depending on the various environmental factors and phenotypic traits such as height, yield, color, and biomass. The key phenotypic signature that has been exploited by researchers around the world to measure the timely yield is the rice grains per panicle known as Spikelets per Panicle (SPP). The existing high-throughput methods for grain count are rudimentary and have several limitations, the most significant being the inability to differentiate and separate the overlapping grains and panicles unique to rice crops.In this paper, we propose and describe a novel method to determine the overall rice production for a dataset containing images of drought-affected and controlled rice crops by the identification and quantification of the total number of rice panicles and estimation of the total amount of grains. Devised a faster R-CNN machine learning model for automating and accelerating the process of identification of the panicles and the spikelets within the panicles. We have conducted a comparative study of rice crops grown in two different environmental conditions, which helps validate the theoretical hypothesis of phenotypic variations in the rice crops. The results obtained prove to be beneficial for the problem statement provider, the Indian Agricultural Research Institute to validate their hypothesis and questions formed on laboratory-based observations. But more importantly, this system would a stepping stone for integrating non-invasive techniques for rice phenotyping in the Indian agricultural sector.