DEVELOPING FACE RECOGNITION FOR CAMERA SURVEILLANCE SYSTEMS

Abstract

Face recognition is one of the biometric information processing methods. Applicability is easier and the working range is larger than other biometric information processing, i.e.; fingerprint, iris scanning, signature, etc. A face recognition system for camera surveillance is designed, implemented, and tested in this thesis study. The system performs the human face recognition task via two main steps: detecting the human face from the frame in the camera and identifying the corresponding individual in the dataset. The system detects the human face with Haar Like Feature - Cascade of Boosted Classifiers and Histograms of Oriented Gradients followed which, it extracts features and recognizes the human face region pre-train VGG16 model to identify who is the individual corresponding to the human face region. The numerical test results on a real dataset with 8895 images collected from 12 students from Can Tho University show that the model achieves an accuracy 93.44%. The system has a simple interface for convenient communication.