Zhang Guomei
ABSTRACT
With the rapid development of information technology, facial recognition technology has been widely applied across various fields, particularly in education, where it significantly simplifies payment processes and enhances management efficiency. This study aimed to design and develop an AI-based facial recognition payment system to improve the security and efficiency of campus payments. The system utilized the RetinaFace algorithm for facial detection and the FaceNet model for high-precision facial recognition, ensuring smooth payment processing. The research employs both quantitative and qualitative analyses. Quantitative analysis evaluates the system’s recognition accuracy, processing speed, and payment security using experimental data. Standard datasets were used to assess accuracy and error rates, while load testing was conducted to analyze system performance under high concurrency. In addition, ISO/IEC 25010 software quality standards were applied to quantify the system’s functionality and reliability. The results indicate that the system achieved a facial recognition accuracy rate of 97.8%, with excellent performance in processing speed and security. Qualitative analysis gathered subjective feedback from parents and students through interviews and observations, focusing on the system’s usability and security. The analysis provided suggestions for further system optimization to ensure that performance improvements align with user needs. By integrating quantitative data with qualitative feedback, the study comprehensively evaluated the system’s effectiveness, ensuring that performance standards were met while enhancing user experience. Compared to traditional payment methods and existing facial recognition systems, this system reduces the reliance on physical payment tools and significantly improves payment convenience. Although factors such as obstructions and lighting variations still affect recognition performance, the system demonstrated robustness in handling these challenges. The system also showed rapid response in handling high-concurrency payment requests, indicating its feasibility and adaptability in campus settings.
Keywords: Artificial intelligence, face recognition payment, facenet, retinaface
https://doi.org/10.57180/vpoh3377