Uso de técnicas de análise de dados e visão computacional na deteção de actividades suspeitas durante a realização de exames de admissão caso de estudo: estudantes da Faculdade de Engenharia da UEM

Abstract

This work aimed to use data analysis and computer vision techniques to detect suspicious activities during admission exams at Eduardo Mondlane University. The study focused on the Faculty of Engineering at UEM, with the purpose of developing a system capable of identifying fraudulent behaviors. Through a literature review, the fundamental concepts of computer vision, image processing, and machine learning, which form the basis for detecting suspicious activities using computer vision techniques, were initially presented. Subsequently, the steps, approaches, and algorithms used for video and image data analysis were detailed. In the case study, a prototype system was created that uses computer vision and data analysis to automate the process of identifying unauthorized behaviors in admission exams. In the developed system, the use of machine learning algorithms for classifying behavior patterns and detecting anomalies, such as the use of electronic devices or communication between candidates, is highlighted. For testing purposes, videos and images manually captured to simulate the admission exam environment were analyzed, and the system achieved a relevant success rate in identifying suspicious activities, correctly classifying a high percentage of cases. Thus, it was concluded that the implementation of a system using computer vision is a solution to minimize the occurrence of fraud during admission exams.