Facial recognition using computer vision techniques

• Main Author: Chow, Jane Kiat Ying
• Other Authors: Chua Chin Seng
• Format: Final Year Project
• Language: English
• Published: 2018
• Subjects: DRNTU::Engineering::Electrical and electronic engineering
• Online Access: http://hdl.handle.net/10356/74788
• Institution: Nanyang Technological University

In this project, face recognition technology is incorporated into webcams to empower traditional security for online transactions and ATM-related frauds. Face detection and recognition are done in real time and in uncontrolled environments to simulate real life conditions. This report aims to evaluate the accuracy and limitations of OpenCV’s Local Binary Patterns Histograms (LBPH) algorithm in facial recognition by determining factors that are significant in affecting the accuracy of the algorithm. A guess of significant factors is determined by an initial experiment. The results of this experiment are then taken and normalized before being used as a comparative benchmark against subsequent tests that are set up. A series of tests are created, each with differing image database with the purpose of testing for significance of the particular factor. The rationale behind this is that if a factor is significant, having images of the factor will improve the accuracy of the algorithm. In the same line of logic, if the factor is insignificant, no matter how many images from the factor image set is in the database, the resulting confidence level for tests will remain constant. Significance is also determined by confidence level produced by the tests, where the confidence level of tests for a certain factor has to be lower than default tests without the factor’s image in the database. After the significant factors are determined, a test with a combination of all the significant factors will determine the optimum threshold for confidence level thereby optimizing the recognition algorithm.