Novel book check-­out system based on Facial Recognition and RFID techniques

Student: Balaji Sampathkumar, 2017-2018

The goal of this project is to create an efficient, commercially viable Facial Recognition-based book check-out system for University Libraries. The Book Check-out system, described herein, will be coupled with an RFID transponder system, which can detect multiple books (qty: 5 – 10) at once thereby reducing a patron’s check-out time from about 50s (assuming a student borrows 5 books) to approximately 4s. Artificial Intelligence and Machine Learning have evolved significantly over recent years and have led to the creation of algorithms, based on Deep Neural Networks, that have a high accuracy rate under optimum conditions. Key factors that adversely affect the accuracy of these integrated systems are low lighting conditions and huge databases (accuracy drops from 93% to 80% under poor lighting conditions and with millions of images in the database). These concerns, however; do not apply to this project, since Libraries are well illuminated and the required algorithms will be derived for relatively small, finite databases (just around 12,000 faces). Further, facial image databases will be populated and updated annually using images captured when patrons apply for their university id cards. Finally, recognizing that transaction security is a growing concern for nearly every business, it is anticipated that this Facial Recognition System will be extensible to other markets and facilities such as banks, weapon storage facilities, hospitals, and grant-access systems in all high-security buildings, and prison to name just a few.