Hui Yu Shen
Chienkuo Technology University Department of Information and Network Communication Associate Professor
Email：wyshen@ctu.edu.tw
Yeong-Chin Chen
Asia University Graduate School of Computer Science and Information Engineer Professor
Email：ycchenster@gmail.com
JHANG, JIN SIAN
Asia University Graduate School of Computer Science and Information Engineer Student
Email：107121007@gm.asia.edu.tw
Abstract
In this research, an AIoT (AI & IoT) system platform is established. By using IoT (Internet of Things) technology, we combined the Raspberry Pi with RFID, LCD and face recognition technology together to collect student ID data and student identification results. Additionally, we used MQTT to send student ID data and student face recognition results to the back-end Big Data server. The big data platform technology used in the back-end includes Apache Kafka for the use of publish / subscribe functions for student face photos in Apache Hadoop HDFS storage. At the same time, the student ID data and identification results are published to MongoDB storage, if the recognition result is lower than expected, then transfer learning is triggered to retrain the artificial intelligence model.
By way of the retraining mechanism of the student’s face recognition model designed by the above technology, three different data processing processes based on the results of student’s face recognition are implemented. First, if the recognition result is higher than the preset value of retraining, only the relevant roll call records are stored; Second, retrain and update the trained model to the front-end Raspberry Pi if the recognition result is located between retraining preset value and recognition low bound; Finally, if the recognition result is lower than the recognition low bound, the back-end system will send an email to the classroom management personnel to warn and notify that the RFID data and the recognition results are inconsistent. In such a way, the automation recognition is achieved.
Keywords :Apache Hadoop, Apache Kafka, Apache NiFi, Transfer Learning, MongoDB