Home surveillance in general

Gue, Kai Kit (2025) Home surveillance in general. Final Year Project, UTAR.

Preview

PDF Download (2254Kb) | Preview

Abstract

This project addresses a growing concern in modern society – home security. With increasing incidents of property crime and unauthorized intrusions, there is a rising demand for intelligent surveillance systems that go beyond the limitations of conventional CCTV setups, which often struggle with false alarms and require manual supervision. This project proposes a smart home surveillance system that combines real-time object detection with violence recognition by leveraging state-of-the-art deep learning techniques. The system uses the YOLO (You Only Look Once) framework to detect the presence of weapons, offering rapid identification of potential threats. Simultaneously, a ResNet50-based Convolutional Neural Network (CNN) combined with a Long Short-Term Memory (LSTM) network is employed to recognize violent actions over time, such as assaults or robberies, using temporal video frame analysis. When a human is detected in the scene, these detection modules are triggered to identify weapons or violent movements. If a threat is confirmed, the system issues an immediate alert to property owners or security personnel, enabling quick intervention. By integrating real-time weapon and violence detection in a multi-threaded monitoring system, this solution enhances home surveillance effectiveness and responsiveness, aiming to create a safer and smarter living environment.

Actions (login required)