Designing a Prototype Intrusion Detection System Based on YOLO V5 at Surabaya Aviation Polytechnic

Abstract

Surabaya Aviation Polytechnic requires a more efficient and responsive surveillance system
to maintain campus security, particularly in perimeter areas prone to unauthorized access.
Currently, the surveillance system relies on conventional CCTV cameras that require manual
monitoring by security personnel, resulting in limitations in detecting incidents in real-time. This
research aims to develop a prototype of an automatic intrusion detection system based on cameras
and the YOLOv5 object detection algorithm. The system is designed to directly detect the presence
of humans and non-human objects through video footage from the cameras. Detected objects are
marked using bounding boxes, and the system distinguishes whether the object is human or not. If
the detected object is human, the system activates a buzzer via an ESP32 microcontroller as a
warning notification. This research adopts a prototype approach, including stages of needs
identification, system design, hardware and software development, and functional system testing.
Testing was conducted using a dataset containing 500 human images and 500 non-human images
to train the detection model. The test results show that the system can automatically detect
intrusions, but it still has limitations in terms of accuracy, detection speed, and notification
effectiveness. Therefore, the system still requires further development.