| Resumo: | This dissertation implements a monitoring system that can be used in three different contexts: monitoring a single student, monitoring a classroom or monitoring a group of people. In order to build this system, we based the development on the use of algorithms for face detection, face recognition and facial features extraction. During this work it was also implemented an eye tracker, a face tracker, an head estimation pose and emotion detection. For each context, different approaches were developed. For the single user monitoring, there was the need of recognizing the face. For this context, the most convenient algorithm was based on Local Binary Patterns Histograms. After a successful recognition, the system assigns an ID to the face and starts tracking it while retrieving useful data from the facial features. For the classroom monitoring, there was no need of recognition, only face tracking. For each face detected, an ID is assigned and the face tracker starts. For the monitoring of a group of people, there was the need of making a face recognition each time a new face appears in the frame and, after a successful recognition, the face tracker starts. The main contributions of this thesis are an automatic calibration for the digital camera used in the system for a better face recognition, a modular solution separated in three components that can be used to monitor three different contexts, retrieving relevant information during a certain period of time that an individual was in front of the camera. The developed software was integrated in a graphical user interface software provided by the camera manufacturer.
|
|---|