Desarrollo de un prototipo de una aplicación móvil híbrida de reconocimiento facial para la DINASED de la ciudad de Riobamba.

Abstract

In the current graduation work a prototype of a hybrid mobile application of facial recognition was developed, Android Studio and Opencv library allowed the implementation of facial recognition algorithm EBGM, a web module was developed in java language, using boostrap, javascript and ajax, this aims to manage cases within the “Dirección Nacional de Delitos Contra la Vida, Muertes Violentas, Extorsión y Secuestros” (DINASED). For the development of the prototype and the web module the SCRUM methodology was applied, which allowed to plan the delivery of the requirements according to a priority, using the T-Shirt estimation method and the time established by the client. For the tests carried out with the mobile prototype, three databases were used: ORL face, Normalized Yale Face and one of its own, taking into account the variation of light, distance, accessories and facial expressions. With the application of the algorithm an acceptance percentage around 25% and 50% between the three databases was obtained, applying a total of 46 tests with jpg images format, rgb scales or for gray with the approaches previously detailed. Regarding the behavior over time and the use of resources that are part of the efficiency, the first one improved by 70%. Therefore, it is possible to perform a greater amount of recognition tests of a person in a time interval, the second one takes into account the amount of images and devices that are used to carry out the recognition process, improving this process by 4 and 8 respectively. The prototype is functional in 92%, it is recommended to take into account the stated approaches in this research because if they are applied correctly, it will allow obtaining quality results in a more efficient way.