Elaboração de um banco de imagens para a calibragem de um sistema semiautomático de contagem de ovos de Aedes aegypti

Abstract

Entomological surveillance is an important strategy to know the occurrence of Aedes aegypti in the environment, and plan actions to control this mosquito. For this, one of the tools that can be adopted are the ovitraps, which are sensitive instruments for detecting A. aegypti in the environment, in addition to having a low cost and requiring little maintenance. However, counting the eggs obtained through these traps has been a laborious job, since each oviposition substrate can contain hundreds or thousands of eggs, which are manually counted with the aid of a magnifying glass by a trained professional. Operator fatigue and limited visualization of the substrate regions by the magnifying glass, as well as egg overlap, are factors that can lead to counting errors and make obtaining entomological data more time consuming. With that in mind, egg counting techniques from images have been studied, aiming to optimize the work, but still in an incipient way. Thus, this work sought to establish criteria for egg counting through image processing. For this, an image bank was created, with 40 specimens, of two types of oviposition substrates, one in wood (M) and another in fabric (T), having different densities of eggs. These substrates were submitted to manual counting and later their images were used for tests in digital processing. As a result, we observed better responses to processing on tissue substrates (T), with no statistical difference (p=0.1091) between manual and semi-automatic counting, whereas the results for the group (M) showed a statistical difference (0.00133 ) among the counting methods, verified through the Mann-Whitney statistical test. In addition, we were able to establish criteria for possible improvements in obtaining images that may favor processing. We therefore consider that establishing better conditions of focus, lighting and drying of the oviposition substrates to obtain the images can considerably improve the result for the two materials worked, requiring further tests.