In 2018, the Epidemiological Record 36 of the World Health Organization (WHO) indicates that around 390 million people get infected from Arboviroses or mosquito-borne diseases. Among the transmission vectors of these diseases, the Aedes aegypti and Aedes albopictus are responsible for a considerable parcel of the infections since they can transmit a broad range of infections (e.g., dengue, yellow fever, and chikungunya). To reduce the number of infections and deaths caused by these mosquitoes, monitor and control the population of these insects is a key factor. In this sense, ovitraps can be employed to monitor the population of Aedes mosquitoes. Ovitrap is a dark container filled with water where a porous wooden paddle is inserted to serve as an oviposition substrate. These devices are installed in monitored areas and, periodically technicians collect them to count the number of eggs deposited in the paddles manually. Because the manual egg counting task can be time-consuming and susceptible to human errors, in this work we present a solution that uses deep learning algorithms to automate the counting process. Moreover, to further reduce the human effort in the counting process, hardware that automatically acquires the images of the wooden paddles is also presented. Experiments comparing the proposed solution, the manual counting method, and two other solutions, namely ICount and EggCounter, are performed. The results achieved indicate that the proposed method achieved a superior result than the two other methods. Moreover, the application of the Wilcoxon test with a confidence interval of 95% indicates that the solution presented can be as accurate as of the manual counting method which is currently adopted.
Read full abstract