Identifying Aedes aegypti Mosquitoes by Sensors and One-Class Classifiers

Author(s)

Souza, Vinicius M. A. ; BeltranCastanon, C ; Nystrom, I ; Famili, F

• Document type: Journal article
• Source: PROGRESS IN PATTERN RECOGNITION, IMAGE ANALYSIS, COMPUTER VISION, AND APPLICATIONS, CIARP 2016; v. 10125, p. 9-pg., 2017-01-01.

Abstract

Yellow fever, zika, and dengue are some examples of arboviruses transmitted to the humans by the Aedes aegypti mosquitoes. The efforts to curb the transmission of these viral diseases are focused on the vector control. However, without the knowledge of the exact location of the insects with a reduced time delay, the use of techniques as chemical control becomes costly and inefficient. Recently, an optical sensor was proposed to gather real-time information about the spatio-temporal distributions of insects, supporting different vector control techniques. In field conditions, the assumption of knowledge of all classes of the problem, it is hard to be fulfilled. For this reason, we address the problem of insect classification by one-class classifiers, where the learning is performed only with positive examples (target class). In our experiments, we identify Aedes aegypti mosquitos with an AUC = 0.87. (AU)

• FAPESP's process: 11/17698-5 - Classification of non-stationary data stream with application in sensors for insect identification
• Grantee: Vinícius Mourão Alves de Souza
• Support Opportunities: Scholarships in Brazil - Doctorate
• FAPESP's process: 15/16004-0 - Intelligent tools to vector control and population orientation against dengue fever
• Grantee: Vinícius Mourão Alves de Souza
• Support Opportunities: Research Grants - Innovative Research in Small Business - PIPE