Smart Campus Monitoring Based Video Surveillance using Haar Like Features and K-Nearest Neighbour

Abstract

Intelligent video surveillance systems based on Internet of Things (IoT) technology have proven to be major primary components for security in many areas, such as smart cities. These systems are important because they provide messages that transfer information about people on campus among camera nodes, thereby providing real-time video surveillance monitoring. The proposed system consists of several cameras and an intelligent processing system, represented by a raspberry pi. The cameras are distributed in different locations in the university campus. Each camera node is connected to the internet and can communicate and share information with other nodes, as well as communicate with a central monitoring (server) via Message Queuing Telemetry Transport (MQTT) IoT protocol. The cameras can extract information in real time from video, and identify everyone as either students, teachers and/or employees using computer vision algorithms. Two methods of face detection and recognition techniques are applied: a feature-based technique that uses the Haar cascade, and an image-based technique that uses k-nearest neighbour (kNN). Face detection and recognition based on the Haar cascade classifier is more suitable for resources with embedded limited systems since it requires less computation, while kNN is more accurate and shows better results in a dynamic environment. All programs were written using open-source Python under a Linux operating system and by using OpenCV library.