Development of a Tool for Automatic Counting and Estimation of the Hatchability Percentage of Rhipicephalus microplus Tick Eggs Using Image Processing and Deep Learning

Abstract

The control of bovine ticks, an ectoparasite that causes significant losses in Brazilian and world cattle farming, is commonly carried out using chemical tickcides, whose efficacy is determined by laboratory tests. The most commonly used test is the Adult Immersion Test (AIT), in which fed females are treated with the tick killer and their reproductive parameters are assessed to determine the product’s effectiveness. However, the AIT is an analysis method that requires a long time to wait for the results. This paper proposes a tool to automate a crucial part of the AIT, the stage of determining the hatchability percentage of tick eggs, using image processing and deep learning techniques. The tool uses the DenseNet convolutional neural network (CNN) to classify the eggs and estimate the hatching rate. For egg classification, the CNN showed a correct classification rate for fertile eggs of 98%, and for infertile eggs 90.3%. The hatchability rate, obtained by counting fertile and infertile eggs, showed an average accuracy of 93%. This approach provides a considerable reduction in the time needed to obtain the results, representing a significant advance in the efficiency of bovine tick control.