Computer Vision System for Automatic Identification of Potential Aedes aegypti Mosquito Breeding Sites Using Drones

Author(s)

Gustavo A. Lima ^[1] ; Rafael O. Cotrin ^[2] ; Peterson A. Belan ^[3] ; Sidnei A. de Araújo ^[4]

Affiliation

^[1] Universidade Nove de Julho. São Paulo - Brasil

^[2] Universidade Nove de Julho. São Paulo - Brasil

^[3] Universidade Nove de Julho. São Paulo - Brasil

^[4] Universidade Nove de Julho. São Paulo - Brasil

• Document type: Journal article
• Source: RISTI - Revista Ibérica de Sistemas e Tecnologias de Informação; n. 43, p. 93-109, 2021-09-30.

Abstract

Abstract Drones have become an important technological tool to help fight mosquito breeding sites. However, the images acquired by them are usually analyzed manually, which can consume a lot of time in inspection activities. In this work, a computer vision system (SVC) is proposed for the automatic identification and geolocation of potential breeding sites of the Aedes aegypti mosquito from aerial images acquired by drones. The developed SVC gave rise to a software, whose core is composed of a convolutional neural network (CNN) that presented rates of recall and mAP-50 (mean average precision) of 0.9294 and 0.9362 in the experiments conducted with a database composed by 500 images. These results, compared with recent results from the literature, corroborate the adequacy of the CNN to compose the SVC, which can bring improvements to the use of drones in programs of prevention and combating mosquito breeding sources. (AU)

• FAPESP's process: 19/05748-0 - Computational approach for Automatic identification of possible foci of the mosquito Aedes aegypti from aerial images acquired by UAVs
• Grantee: Sidnei Alves de Araújo
• Support Opportunities: Regular Research Grants