Design and Development of Sugarcane Maturity Identifier through Phenotypes via Image Processing

Abstract

Most of the world's sugar demand comes from sugarcane. Sugarcane is the most produced crop globally and one of the major crops in the Philippines. The Philippines' sugarcane industry shows a decrease in the total annual production. Maturity is one factor that affects yield and, eventually, production. Sugarcane must be harvested at the proper age (peak maturity) to maximize sugar output. Among the different approaches to identify maturity, use the physical and physiological aspects. Approaches effects lead to the design and development of a system that will determine maturity through phenotypes via image processing. The system will process images of the sugarcane varieties of the Philippines, using Raspberry Pi and send/receive them using Long Range Wide Area Network (LoRa WAN). Pythons' object detection algorithm, specifically Faster Region-based Convolutional Neural Network (Faster R-CNN) and pre-trained models in TensorFlow, are used to identify maturity. The results have shown that the system performs well in identifying maturity and has excellent potential to be used in sugarcane production, which eventually increases sugar production.