Development of a real-time automated family classification system

• Main Author: Chen, George Fengrong.
• Other Authors: Teoh Eam Khwang
• Format: Final Year Project
• Language: English
• Published: 2011
• Subjects: DRNTU::Engineering::Computer science and engineering::Computing methodologies::Image processing and computer vision
• Online Access: http://hdl.handle.net/10356/45358
• Institution: Nanyang Technological University

The local binary pattern LBP, which was originally used for texture analysis, is now being proposed for improvement by the addition of threshold information. The LBP is superior in computational simplicity and this allows this project to be implemented in real time family classification system. As such, this project has two main components, one is to access the performance of proposed LBP with thresholds information which is in the back end stage, and the other is to implement this in real time application which is in the front end stage. Accessing the performance of the thresholded LBP, in terms of its error rate and training time was experimented in this project. Experimental results showed that the thresholded LBP, with thresholds from -69 to +69, step size of 5, has an improved error rate of about 4% and the most significant improvements was 8% in the family classification experiment of 53x63 pictures.