Design of epidemic detection system in railway station based on FPGA

Since the outbreak of COVID-19 in 2020, wearing masks and displaying green codes in and out of public places has become a habit. The identity verification of the existing railway station is mainly based on face recognition. Due to the outbreak and persistence of the epidemic, people often need to call out the green code on their mobile phones for staff verification, which takes a lot of time. At the same time, the existing face recognition equipment requires the inbound personnel to take off their masks, which also increases the infection risk of the inbound personnel. In order to reduce the above infection risk and speed up people’s entry and exit speed, we have designed a system that can identify people wearing masks and judge whether they are confirmed or suspected cases at the same time. Firstly, the system measures the passenger’s body temperature through the infrared temperature measurement module, carries out face detection and recognition at the same time, and queries the recognition results in the database to judge whether the passenger is diagnosed or in close contact. When the passenger is normal, it is allowed to pass, otherwise it is not allowed to pass, and updates the relevant data in the cloud database. The system uses Yolo algorithm as the face detection algorithm, and then carries out face recognition through FaceNet network, so as to judge its identity and query the relevant information of the person in the cloud database. After testing, the iterative loss rate of the system is basically below 0.1 and the accuracy is basically stable above 99%. Considering that we need to use it on embedded devices and the amount of calculation operation of deep learning algorithm is large, and FPGA can well build circuits according to the needs of the model because of its reconfigurability, and FPGA can realize hardware acceleration because it can run in parallel, so we finally choose to deploy the model to FPGA to complete face recognition.