AI-Empowered Intruder Alert System for Enhanced Security Surveillance

In today’s contemporary society, ensuring security and safety is of paramount importance. For a nation to maintain a robust economy, it must prioritize the establishment of a secure environment. Surveillance cameras are widely employed to monitor and detect illicit actions. However, the current approach often relies on human intervention and monitoring, primarily after an intrusion has occurred. To address this issue, an automated system capable of identifying illegal activities without constant human supervision is necessary. This will help local security. Current local retailers of CCTV systems in Kerala lack AI-enabled intruder alert systems. The proposed project focuses on developing a software to integrate with the current software of the CCTV companies, which provides customizable intruder detection solutions. This will help the local security companies sustain themselves against international competitors by applying machine learning algorithms that take real-time video as input and detect the presence of intruders on the premises and alerts the user/admin. A machine learning model will be built using Convolutional Neural Networks and trained using a large dataset of real life CCTV footage as inputs. The model will be trained to recognize humans with or without weapons, animals, and unidentified moving objects separately and specify the same in the alert provided and also specify familiar faces or not. Users can specify a certain alert time during which the intruder detection system will be activated. This model will be cross-platform integratable, i.e, can be integrated with a mobile/desktop/web application as per the requirement of the client. Accuracy, Precision, Recall and F-measure has been computed for the proposed model.