MUCET 2021: Implementation of Multiple Face Detection for Surveillance Systems
MUCET 2021 (123 P7-Jebat)-16/11/2021 (8.40 – 8.80 am)
Paper Title: Implementation of Multiple Face Detection for Surveillance Systems
Prepared by: Dr Wira Hidayat bin Mohd Saad (https://blog.utem.edu.my/wira_yugi/)
(MLSP. Computer Department, FKEKK, UTeM, Durian Tunggal, Melaka, Malaysia)
Prepared for: MUCET 2021 (https://mucet2021.utem.edu.my/)
Full paper: https://crim.utem.edu.my/wp-content/uploads/2024/05/186-378-3791.pdf
Abstract
Surveillance systems such as closed-circuit television (CCTV) using pan-tilt-zoom (PTZ) cameras can change the position of the scene by pan, tilt and zoom features. The downside of the camera is unable to capture the face clearly as it focuses on the static point of the lens. Thus, the purpose of the project is to develop a surveillance system that is capable of detecting faces within the camera view. This project was developed using LabVIEW and Microsoft Visual Studio. The data from the camera and navigation system were implemented by using LabVIEW, whereas the face detection system was implemented by using both LabVIEW and Microsoft Visual Studio. The performance of the face detection was examined by conducting various kinds of experiments, which included facial orientation, movement of the face, obstruction on the face, and light conditions. The accuracy was achieved at 89%, and the system was able to detect a minimum of 10 faces in a view of scene
Summary
As someone who’s worked with surveillance systems, I’ve always found it frustrating that conventional CCTV setups—even those equipped with fancy pan-tilt-zoom (PTZ) cameras—often fail at one critical job: capturing clear, usable images of faces. It’s a common weakness that compromises security and makes post-incident reviews less effective.
So, in our recent research project (which I had the privilege of presenting at the MUCET 2021 conference), my team and I set out to build something smarter.
We developed a multiple face detection system using a PTZ camera integrated with LabVIEW and Microsoft Visual Studio, applying the well-known Viola-Jones algorithm. The result? A real-time system that can detect up to 10 faces in a single camera view—even under tricky conditions like poor lighting, partial face obstruction, or varying orientations. The best part? We achieved an impressive 89% accuracy rate in our tests.
But this is just the beginning.
Our work lays the foundation for more advanced capabilities like auto-tracking and face recognition—the kind of intelligent surveillance you’d expect from high-end security systems, but made more accessible through smart engineering.
This post is a quick glimpse into that journey—how we designed the system, why it matters, and where this technology could go next. If you’re curious about the future of surveillance, or just love seeing how simple ideas evolve into working systems, I invite you to dive in.