Towards automatic classification of mosquito species based on wing geometrical features

This paper presents an initial version of a system for the automated classification of mosquitoes species, based on relevant features extracted from their wing's morphology. The algorithm developed allows identifying the mosquito's species by using key reference points of the wing, such as the radio of the circular geometries of spots presents within the wing. The aim was to develop an initial version of a system for improving the standard manual method in which mosquitoes are classified, as a proof of concept. For testing the system, two particular species: Limatus durhamii and Wyeomyia sp. were used for classification using a simple perceptron. The model reached an accuracy value of 95.46% in predicting new wing samples. Initial results indicate that with future refinements, an automated classification system is feasible.