Vision-based security system using two-view IP cameras

• Main Authors: Cansado, Jerome C., Hatol, Christian Glenn T., Primavera, Jhonas N.
• Format: text
• Language: English
• Published: Animo Repository 2016
• Subjects: Security systems; Engineering
• Online Access: https://animorepository.dlsu.edu.ph/etd_bachelors/8748
• Institution: De La Salle University

Maintaining the security in large facilities like a university is one of the most difficult tasks to do. The use of the Vision-Based Security System Using Two-View IP Camera is one of the methods in strengthening the security in De La Salle University Science and Technology Complex. The purpose of this thesis is to increase the security of DLSU-STC through the recognition of the people who enters through the East Canopy of Milagros R. Del Rosario Building. The system uses two internet protocol (IP) cameras one for the front view and one for the right profile view, Linear Binary Patterns (LBP) for the face detection from the moving subjects, Support Vector Machine (SVM) for the actual facial recognition, Creator for the graphical user interface, Facial Alignment and Object Annotation for gathering the database, and T-Test for the statistical analysis of the data. The researchers first decided to use Linear Discriminant Analysis (LDA) however the results were underwhelming and decided to use feature extraction using KAZE keypoints and descriptors with SVM instead which provided better results.