Developing a framework for using face recognition in transit payment transactions

Abstract

Nowadays, significant number of people relays on public transportation to commute to their final distention due to the increase of the private cars cost, traffic jam, toll gates, high petrol charges and other factors, which create a huge pressure on the public transportion infrastructure in general and the fare collection system in specific. Therefore, transit operators are continuously keen to identify different solutions to reduce that pressure and improve the travel experience by upgrading its fare collection system to the advanced state-of-the-art account based ticketing system in order to achieve better flexibility to offer smooth and convenient payment options for the passengers to choose. On the other hand, a tremendous advancement has been noticed in the human face detection and recognition technology which mainly used to authenticate and identify person face from a group of people through detecting a unique feature of the face and ignore the background image then compare the outcomes with the registered faces in the database to identify the person. This dissertation proposes a framework which aims to offer face recognition technology as a new payment option inside metro station. The proposed framework involves the hardware, software, algorithms, and system specification requirements. Further, it provides a detailed end-to-end systems integration and transaction flow between the account-based ticketing, face recognition, and banking systems. It’s worth to mention that the proposed framework is built based on the outcomes of three dimensions, including a systematic literature review, users’ surveys, and experts’ surveys. 84% of the users expecting an improvement to their travel experience if the face recognition access offered. In addition, the experts supported the users’ survey results by claiming the optimum technical feasibility to implement the face recognition access inside metro station. The framework offers two state-of-the-art solutions. The first solution is proposed based on integrating the existing surveillance camera systems with the recommended “Banking Payment Context- Account Based Ticketing System” to offer face recognition access entry to the passenger inside metro station. A number of combined algorithms and classifiers are proposed to use in this solution based on the encouraging outcomes observed from the systematic literature review and experts’ survey, including Local Binary Pattern descriptor, Haar-Like Descriptor, Ada Boost, Cascade classifiers, Affine Transformation, Histogram Equalization, Gaussian Filter, Principal Component Analysis which are embedded in OpenCV or MATLAB application. The argued face recognition accuracy between 98%-99.2% and average processing time including metro gate opening time ranges between 1114-1400 milliseconds. This solution considers an effective cost-based solution. The second solution is proposed based on implementing a dedicated full HD face recognition stereo camera system on top of each metro gate and integrate it with the recommended “Banking Payment Context- Account Based Ticketing System” by using the MFcoface face recognition method which results from the systematic literature review and experts’ outcomes. The argued face recognition accuracy ranges between 99.3%-100% and average processing time including metro gate opening time ranges between 200-400 milliseconds. This solution considers an efficient performance-based solution.