Enhancement of Real-Time Face Recognition via Video Surveillance System

Abstract

This project report is about the investigation to enhance the accuracy of the face recognition video surveillance system. Recently, video surveillance with Closed Circuit TV (CCTV) cameras to analyze crowd people activities and real-time face recognition have been widely used.  There are many face recognition systems that use Haar Cascade feature. However, due to the limitations of Haar Cascade, there are many methods and algorithms that have been implemented to enhance the accuracy of face recognition. One of them is the Gray Level Co-occurrence Matrix (GLCM). The hardware and software used in this project are Raspberry Pi, Pi camera, OpenCV library, and Scikit-Image library. The developed system consists of four phases for the operating method, which are data gathering, GLCM image texturizing, train recognizer, and recognition. Thonny Python was used in all of these phases with the OpenCV library and Scikit-Image library applied. All the results of this project were based on face detection, face recognition with face complications, face recognition with distances, and face recognition with different face angles. The results confirmed that the system was reliable with human face detection and face recognition with an accuracy up to 68%.