Caracterização de micro vestígios entomológicos por imagem: aplicação de redes neurais profundas

Abstract

Micro-traces analysis is a new and growing area of research within forensic science and may be vital in solving lots of criminal cases, assisting and accelerating the identification of criminals and areas in which they act. Nowadays the forensic entomology area has been increasingly used for insects characterization, in order to ascertain the origin of loads of narcoticts, money pouches or even corpses. However, this is still a demanding process, mainly because of the fact that specimens are mostly found in pieces. In order to improve this process, this thesis presents the use of a deep neural network for characterization of insects or pieces of it, up to the taxonomic level of family. The application of the system showed an accuracy of 77,77% among five families: Buprestidae, Formicidae, Pentatomidae, Muscidae and Calliphoridae.