The aim of this thesis is to analyze the behaviour of the people passing in front of a target scene. We consider an environment composed of a so-called target scene (a specific scene under analysis, such as a large plasma screen, a projected image, an advertising poster, a shop window, etc.) and a monitored area (place from which people look at the target scene, such as a street or shopping mall). Computer vision provides promising techniques enabling to obtain such information by analyzing videos captured by cameras monitoring this area. Such information are useful in order to simplify technologies that uses the output of the studies about a target scene. In this thesis, we propose an approach that estimates the visual gaze of a person in a controlled environment. The visual gaze of a person is estimated from the head pose. It is followed by its projection on the target scene that allows to estimate the approximate location of interest. Finally, an analysis of the region of interest allows an accurate explanation of the human activity and interest.
