Mobile system for the identification of pest in the lucuma leaf based on artificial intelligence in Villa El Salvador (#1847)

Authors

Huarote Zegarra, Raul Eduardo

Abstract

The present research focuses on applying artificial vision and the artificial neural network, for the identification of pests in lucuma leaves (SP, PG, ER) in the field of the district of Villa El Salvador through a mobile system with 4500 images obtained in the field, managing to do it from a mobile application being easy to use, as well as the results will be reliable, since the results are statistically demonstrated (confusion matrix), where 99.32% of accuracy, an f1-score of 99%, 100% and 0.99% for SP, ER and PG disease respectively. Thus, it was also possible to achieve an effectiveness rate of 99.32%, implying that the adjustments in the configurations necessary for the SOM neural network model are ideal and the algorithms based on computer vision used (sobel, medfield and orientation) are ideal, therefore the first specific objective is met, in the same way it was possible to reduce the pest per crop area by 1.5% (visually) due to the care taken in identifying the pest on the leaf, therefore, it can be said that Considering the limitations of nature, it meets the second specific objective. Considering the level of simplicity of using the Smartphone, where 33% are extremely satisfied, 45% are very satisfied, 22% are somewhat satisfied and 0% are not so satisfied with not at all satisfied.