With the continuous progress of Internet of Things technology, the development of lockers has ushered in new technological changes. However, the current locker still has problems such as a low degree of intelligence, low unlocking security, and inability to remotely and systematically understand the various information states of the locker.

To this end, a AI intelligent face locker is designed, which uses deep learning algorithm (Deep Learning) to realize face recognition, and transmits the recognition result back to the locker control end to realize the switch of the locker. Finally, the information of various sensors built into the locker is uploaded to OneNET cloud platform in real time for remote monitoring of locker information and data statistical analysis.

This work is highly innovative compared to traditional lockers or existing smart locks on the market. First of all, the work uses deep learning algorithm, uses convolutional neural network (CNN) to learn the feature information extracted from the entered face picture, compares its output with the feature of the person to be identified, and calculates the similarity to realize identity recognition. Compared with the traditional face recognition algorithm, it has stronger adaptability, shorter recognition time and higher accuracy. Secondly, STM32 is used as the main control chip of the system to realize sensor signal acquisition and motor lock control, and various sensor data are uploaded to OneNET cloud platform to realize locker cloud monitoring and data statistical analysis. Thirdly, the cabinet is equipped with vibration sensor and buzzer. When the cabinet is found to be abnormal, an alarm will be sounded to prevent malicious lock picking behavior.

After the design of the work was completed, we tested the function of the work. First of all, different faces are used to verify the function of brushing face and opening the cabinet, thus realizing accurate and fast specific face recognition. Secondly, by observing the switch action of the locker after face recognition, the unlocking action of the locker can be automatically completed under the correct face. Thirdly, enter the OneNET interface to view the sensor data in the locker and the daily report to complete the final locker cloud monitoring and data statistical analysis.