DEVELOPMENT OF A SHOPLIFTING DETECTION AND NOTIFICATION SYSTEM USING COMPUTER VISION IN CCTV FOOTAGE

ANI, OBINNA GEORGE (2025) DEVELOPMENT OF A SHOPLIFTING DETECTION AND NOTIFICATION SYSTEM USING COMPUTER VISION IN CCTV FOOTAGE. Other thesis, GODFREY OKOYE UNIVERSITY, ENUGU.

Text REPO286.docx Download (2MB)

Abstract

This study presents the design, implementation, and evaluation of an intelligent violence detection system leveraging a combination of MobileNetV2 and a “Long Short-Term Memory (LSTM) model” for real-time analysis of surveillance video footage. The MobileNetV2 is a pretrained Convolutional Neural Network (CNN) algorithm. The system's Python development and deployment on an edge processing device allow for the effective extraction of temporal and geographical information essential for correctly classifying violent and non-violent behaviors. Before being input into the CNN for feature extraction, video material is preprocessed. The LSTM then analyzes the data to find temporal patterns suggestive of violence. Through an IoT communication node, the device automatically sends out email notifications to authorized stakeholders upon detection, facilitating prompt action in delicate settings, including public gatherings, shopping centers, and schools. The approach's robustness and usefulness are validated by extensive testing on a benchmark dataset, which shows the model's excellent accuracy, precision, and recall with few false alarms. The study emphasizes the system's potential for improving automated surveillance capabilities, albeit with several restrictions on data variety and processing delay. Future research is advised to increase dataset variety, maximize computational effectiveness, and incorporate multi-modal alert systems to enhance scalability and real-world application.

Item Type:	Thesis (Other)
Subjects:	Q Science > QA Mathematics
Divisions:	Faculty of Natural and Applied Sciences
Depositing User:	Uchenna Eneogwe
Date Deposited:	04 Jun 2026 08:46
Last Modified:	04 Jun 2026 08:46
URI:	http://eprints.gouni.edu.ng/id/eprint/5745

Actions (login required)

View Item