Autonomous Camera Alignment System

• Main Author: Salehian, Sina
• Format: Final Year Project
• Language: English
• Published: 2017
• Online Access: http://utpedia.utp.edu.my/19075/1/06.%2017737_Dissertation.pdf; http://utpedia.utp.edu.my/19075/

The most critical step in the process of identification or verification, using iris or facial recognition is the image capturing stage. A high quality image contains more accurate data about the features in an image. In iris or facial recognition system the rate of success depends on the quality and the angle of the camera when capturing the image. The highest rate of success in verification or identification is achieved when a high resolution picture is taken, with the camera positioned exactly in front of the subject’s face, as demonstrated by some of the previous studies done on the topic. Thus, a camera alignment system is required for situations where the subject might move his/her head. The proposed camera alignment system is consist of 2 independent arms that can move to different coordinates in one plane. The subject’s face will first be detected by using the Ada-boost algorithm with Haar cascades, and the width and height of the face will then be translated from pixel to millimeter. The final coordinate of the face can be estimated from the width and the height of the face. Based on the current angle of the arms and the angle that will get the arms to the final position, the arms will be moved to a specific angle. This system requires a powerful embedded platform. Thus, it is fully integrated on a Raspberry Pi.