Artificial intelligence for optimizing pre-harvest processes and measurement of productivity in coffee crops

Abstract

The food production industry and market is undergoing a transformation. Agricultural activities have been operating in a different way from the past, notably due to the technological, scientific and social advances currently achieved. Increasingly accurate sensors, artificial intelligence, digital platforms, gene manipulation, new drugs, professional qualification. All these advances will gradually make this sector more profitable, safer, more efficient and sustainable. From biotechnological advancement to high connectivity, the new digital tools modify and optimize practically all stages of the production cycle: they are capable of making more accurate decisions, continuously monitoring and obtaining data in real time, bringing greater productivity, agility and security coupled with the reduction of costs. costs and sustainability. In the agricultural sector this current period has been called agriculture 4.0. Among the various products marketed by Brazil, coffee is notorious. The coffee bean is one of the most important commodities and Brazil is the main producer and exporter of coffee beans in the world. In the country there are about 2.16 million hectares planted. This sector accounts for almost 10% of national agricultural exports and generates around US$ 5 billion per year. Brazilian coffee farming has faced several difficulties, and among some pre-harvest stages of coffee, a moment is notoriously important: when the fruits are at the point of harvest. Knowing the correct time of harvest minimizes losses of quality in the drink, guaranteeing the coffee grower a better commercialization of the harvested coffee. Currently, in order to evaluate this moment, coffee growers need to carry out manual removal of some plants and quantify the proportion of green fruits, demanding effort and costs. Another point, the harvest estimate is also an activity done in the pre-harvest and important for the coffee grower to plan. Currently, similar to the evaluation for the moment of harvest, the harvest estimate is also carried out by plant shedding, again involving efforts and costs; and as the loss of pellets occurs before the maturation phase, the estimate can be underestimated. Therefore, the aim of this project is to develop a digital tool to assist these three pre-harvest processes: (i) harvest estimate in the pellet phase, (ii) harvest estimate in the fruit phase, and (iii) evaluation for the moment of harvest. (AU)