Sistema de nariz eletrônico aplicado na análise da floração e raleio em pessegueiros

Abstract

The peach tree during its reproductive period presents flowering buds that develop through different stages, such as the opening of the petals and the swelling of the ovary, until the formation of the fruit. To improve the distribution of resources in fruiting plants thinning is employed. The technique reduces the number of buds or fruits, being an activity that brings benefits, but it is costly and in some cases impractical in the entire extension of an orchard. The right moment for execution and the intensity are complex issues that involve specific knowledge

about the conditions of an orchard. Therefore, information that can help planning and decision- making about thinning is relevant and can improve the cost-effectiveness of the technique. In

this study, information about the budding stage was considered the most important, being used to define the general state of development of the peach tree. However, the development of buds occurs simultaneously according to plant characteristics and environmental conditions. One approach to the problem is to analyze peach tree odor and possible variations due to bud development. In this sense, an electronic nose system can be employed, since it is a device that uses an array of gas sensors and pattern recognition to relate odor samples with the purpose of analysis. For this, a prototype was developed and collections were performed during the reproductive period of a peach tree. The research hypothesis was that the electronic nose can identify the general stage of the buds. To verify the hypothesis, a classification analysis was performed. The result showed that all models presented balanced accuracy above 90%, with better performance for models based on K-Nearest Neighbors (KNN), and Random Forest (RF), than those based on Extreme Learning Machine (ELM) and Support Vector Machine (SVM). In the total of 24 models, 11 did not obtain error given the metrics employed. Regarding the research hypothesis, it is considered that the electronic nose was able of distinguishing between the different proposed stages, namely petal fall (QP), initial fruit formation (FFI), advanced fruit formation (FFA), and formed fruit (F). The results of this study indicate a potential use of the electronic nose in aiding decision-making in activities in orchards.