Shopping Mall Smart Surveillance System (Mobile App) – Facial Recognition, Face Mask Detection & Object Classification

Yong, Yung Jun (2022) Shopping Mall Smart Surveillance System (Mobile App) – Facial Recognition, Face Mask Detection & Object Classification. Final Year Project (Bachelor), Tunku Abdul Rahman University College.

Text RDS_Yong Yung Jun_Fulltext.pdf Restricted to Registered users only Download (1MB)

Abstract

The flaws in the current manual security system, which is both ineffective and costly, have resulted in massive losses for shopping mall operators. Visitors' safety in the shopping mall can no longer be guaranteed. As a result, the Shopping Mall Smart Surveillance System is introduced, which includes artificial intelligence features such as facial recognition, face mask detection, and object classification. Facial recognition may be used to identify wanted criminals for police, whilst face mask detection can ensure that all entrants wear masks when entering the shopping mall. To maintain the safety of the shopping mall, object classification is utilized to detect dangerous weapons such as scissors and deter them from entering. CCTV video streams are displayed in a local host website using Python Flask, and data is transferred to Firebase to be synchronized with the content in the mobile app. The mobile app is also used to alert guards in the event of a criminal detection. Unit testing, integration testing, and system testing are all performed to ensure that the system fits the specifications. All of the above-mentioned goals have been met in this project, and the system is ready for deployment. However, the system developed has some limitations, such as an 8-second delay in the CCTV video stream due to Raspberry Pi RAM constraints. The usage of a pre-trained model also limits the model's ability to be changed and renders it rigid. These issues can be addressed in the future by developing one's own model and deploying the artificial intelligence module in the cloud to reduce the computing power on the Raspberry Pi