Segmentação, rastreamento de objetos e detecção de eventos primitivos com aplicação no monitoramento automático de ações humanas em vídeo.

Abstract

Human motion analysis of human movement in digital video has been a frequent topic of research in computer vision due to the large number of promising applications, which are part of a new field, commonly referred to as "looking-at-people". The general objectives of this class of applications are usually to discover the location and identity of humans in the scene, as well as what are their activities on the environment. Within this context, video surveillance systems are ever more gaining popularity because of the large amount of (operator assisted) monitoring systems installed in public places such as shopping malls, sports centers, airports, train stations, public roads, among others. For the design of an automatic surveillance system it is possible to identify three major research challenges, which were the object of study in this work: motion detection, tracking and detection or classification of events according to the application. This dissertation includes a review of computational techniques needed to develop a prototype of a system for detecting the removal of unattended objects in environments monitored by surveillance cameras. This prototype has been implemented and promising experimental results were achieved.