الفهرس 
IntroductionOnly 14 pages are availabe for public view
 |  | from | 133 |  |  |
| | | from | 133 | | |
Abstract
For many applications in the current industry, real-time head pose estimation is a critical question adding to it the relative motion of the capturing camera where applications such as autonomous driving, in which the driver is supported in monitoring his concentration and drowsiness while driving. Another is when using tracking drones or augmented reality applications where the human pose information is mandatory for the visualization of data correctly to the intended person.
The camera can be located in a non stationary location, This research focus on an important topic in this domain of which is the 3D head pose Tracking from a movable camera that provides real-time operation on low-cost boards 10Watt SBC, which can help potential work in recognizing gestures and facial impressions. In order to detect the focus point and the attention of the monitored individual, this information is required to compensate for his motion to take the correct action accordingly.
This proposed framework incorporates many Major computer vision techniques that traditional image processing techniques for image projection, feature detection, key point description, Image matching, RANSAC, Image Homoghraphy and 3D pose estimation along with Machine learning and convolution Neural network techniques for Face Detection, Facial Landmarks localization as well as Kalman Filter for tracking.
This research also present a custom created model for Landmarks localization that supports the system to provide robust and real-time operation for the whole system
xiv
xv
taking into account the limitations of utilizing off the shelf low-cost hardware capabilities and low power consumption needed to the prospective applications. The system only utilizes the data provided through the IR camera sensor to estimate both the Head and camera pose without any need for additional sensors.