Low-Resolution Face Recognition for Real-Time Smart Classroom Check-In: A Novel Framework in Intelligent Computing

This study proposes and improves a real-time sign-in framework for smart classrooms based on a low-resolution face recognition model. This framework can solve the challenges of low efficiency and easy false sign-in in traditional classroom sign-in methods. The framework consists of two core parts: 1. An efficient real-time image acquisition system; 2. An accurate face recognition model in a low-resolution background. At the image acquisition level, this study designed a high-speed image acquisition system. The system uses the Phoenix 6.3 MP CMOS camera from Lucid Vision Labs. At the same time, a hardware-level bilateral filtering algorithm is used to reduce image noise. This operation can improve image quality and ensure real-time performance. In terms of face recognition, this study proposes a low-resolution face recognition model based on deep learning. The model uses an unsupervised domain adaptive transfer learning framework based on a deep neural network architecture. The framework can effectively extract facial features from low-quality images and perform high-precision matching. In the data set experiment, the recognition accuracy of this model on the excellent quality image set reached 99.76%, and on the poor quality image set, it can still maintain an accuracy of more than 90%. In the comparative experiment level, compared with the traditional CNN and FCM algorithms, the accuracy of the algorithm in this study is far ahead.