Abordagens computacionais para análise da resistência de carrapatos bovinos aos acaricidas baseadas em imagens

Abstract

The cattle tick is one of the main threats to the livestock industry. Annually, the action of these ticks, either by disease transmission or by sucking animal blood, causes a loss of 3.4 billion dollars anually in Brazil. Seeking to maintain control of the tick population, most of the people use acaricides to minimize this impact. To choose the best acaricide, product efficacy tests are performed. These tests are done manually. Samples of tick larvae are subjected to these acaricides, and after, the live and dead larvae are visually counted. Over this rate, the effectiveness of acaricide is defined. Although, This test is time-consuming, repetitive and tiring. Thus, the work developed in this project aims to automate the counting procedure. To achieve this goal, 3 automatic counting methods are developed that follow the flow: Image acquisition, larval segmentation, and larval tracking. Each method uses a different segmentation technique. Illumination compensation, Firefly Algorithm, and U-Net are the segmentation techniques used in this work. The larvae are tracked using a circumference detection technique, the Fast Radial Symmetry Transform (FRST). The best overall result was obtained with U-Net, however all methods achieved good results. The proposed method indicated that 98.86% of tick larvae were found, and 99.25% of living and 97.92% dead larvae were counted correctly.