TÉCNICAS DE VISÃO COMPUTACIONAL PARA CONTAGEM DE OVOS DE NEMATOIDES-DAS-GALHAS (Meloidogyne spp.)

Abstract

The automation of repetitive activities has been used for a long time by man, both to reduce activity time and to improve their performance or performance, and the use of computers meets this objective. One of the main activities of a computer is the ease of doing calculations, however, when it comes to accounts that require human skills, such as vision, for example, deterministic algorithms are no longer sufficient, requiring stochastic algorithms of artificial intelligence. The process of recognition and counting of nematode eggs is necessary to assist in the study of the population of this pathogen. Egg counting is performed manually and takes time, as well as care to reduce counting errors. The use of computer vision techniques meets this need, automating and speeding up the recognition and counting of root-knot nematode eggs. Therefore, the objective of this work is to apply computer vision techniques in order to identify and quantify eggs of nematodes of the genus Meloidogyne spp., in images. In this work, a system for egg recognition and counting of root-knot nematodes using convolutional neural networks was developed. The system receives images as input and outputs the number of eggs identified. Accuracy and precision are also evaluated by running the system on a set of test images. The results achieved presented an accuracy of 67% and a precision of 92%. Thus, it was possible to verify that the use of computer vision techniques can provide the counting of nematode eggs in images.