Does the supermarket turnstile support facial recognition access?

一,Technical principle: 3D modelling and live detection are both guaranteed.
Dynamic facial recognition technology is at the heart of the turnstile face recognition system. Its technical structure may be broken down into three levels:
1. Data acquisition layer: Multi-spectral imaging gets around environmental limits
Old-fashioned 2D facial identification doesn't work well in low light or when something is blocking the view. Most new turnstiles use 3D structured light cameras or binocular stereo vision technologies. Shenzhen Chengwei Technology's full-height turnstile, for instance, has an Intel RealSense depth camera built in. A 3D representation of the face with millimeter-level accuracy can be made in 0.2 seconds by combining infrared light spot array projection and visible light photos. This technology not only fixes the problems of reflections and losing face features in bright light, but it also protects against photo and video attacks.

2. Algorithm processing layer: feature extraction based on deep learning
For feature extraction, the system uses a Convolutional Neural Network (CNN). FaceNet, for example, uses the Inception ResNet network to make 128-dimensional feature vectors. The speed gate system at Shanghai Weiliang Electronics uses a Triplet Loss function to match the pictures taken on site with the ID card photo library of the Ministry of Public Security. The gate opens when the cosine similarity goes above 0.7. To fix the problem of poor recognition rates for people with dark complexion, some companies have used the Racial Faces in the Wild dataset to improve their algorithms. This has lowered the false recognition rate from 5% to less than 0.3%.

3. Security layer: data encryption and liveness detection
The device has an infrared liveness detection module that uses the changes in skin reflection spectra to tell the difference between actual faces and 3D masks. The national security SM4 algorithm is utilised to encrypt and store the feature vectors, and the TLS 1.3 protocol is used to send the data. For instance, the smart terminal used by ICBC's Wuhan branch adds micro expression recognition to the facial comparison process. This checks the authenticity of living organisms even further by looking at more than 20 physiological signs, like how often they blink and how much their mouths twitch.

二,System Architecture: Modular Design for Easy Deployment
The current rotary gate face recognition system has a layered architecture with three levels: the hardware layer, the driver layer, and the application layer.

1. Hardware layer: Industrial design with high reliability
Body of the gate: The new rotary gate's box structure is made of 304 stainless steel, and the box may be as thick as 2mm and sustain an impact force of 1000N.
Module for identification: It has an 8-megapixel wide-angle camera and an LED fill light built in. It can work in temperatures from -40 to 80 degrees Celsius.
Control unit: has an ARM Cortex-A53 quad core CPU that can process data at 1.5 TOPS, which means it can finish comparing in 0.5 seconds.
2. Driver layer: an interface that works with many protocols
The system has a number of interfaces, like Wiegand 26/34, RS485, TCP/IP, and others, that may work together without any problems.

The payment system was connected to the Alipay "Dragonfly" face brushing terminal, and the whole process of "identification payment opening" took only 1.2 seconds.
System for visitors: Use the SDK to connect to enterprise OA, create a dynamic QR code after a guest makes a reservation, and then scan the code at the gate to finish face recognition.
System for fire protection: The gate will automatically unlock and stay open when the smoke sensor goes off, as required by GB 16806-2006 fire safety rules.
3. Application layer: Smart management platform
There are three main things that the backend management system does:

Permission management: lets you restrict who may see your whitelist (workers), greylist (VIPs), and blacklist (dishonest people) in a graded way.
Data analysis: Statistics on the flow of passengers and the distribution of passage periods in real time to help supermarkets run their businesses.
Remote operation and maintenance: Using 4G modules to send out fault warnings and parameter distributions to lower the cost of maintenance on site.
三, Use case: a full overhaul that includes everything from making things more efficient to keeping them safe and in control
1. A store without people
The revolving gate technology at Hema Fresh's Shanghai Jinqiao store creates a closed loop of "face recognition in store shopping face recognition out." For the first registration, users must upload a picture of their ID card. After that, it simply takes 0.3 seconds to log in. The technology connects to ERP, automatically takes points away from members, or ties to an Alipay account. This makes the daily average traffic efficiency of a single store three times better.
2. Situations for stopping and controlling an epidemic
When the body temperature is found to be 37.3 °C or higher, the gate integrated infrared temperature monitoring module employed by Wuhan Zhongbai Warehouse instantly sounds an alarm and refuses to open the gate. During the Wuhan outbreak in 2022, the system examined 12 people per day who were not normal, which effectively stopped the spread of the disease.
3. Situations for managing loss prevention
Through turnstiles, Yonghui Supermarket Fuzhou Taihe Plaza store keeps track of when each consumer comes and goes. It also uses video surveillance to look at the shopping trail. The loss rate for the goods went down from 1.2% to 0.4% after the system was put in place. This saved the company more than 2 million yuan a year.
Painted Pedestrian Controlled Speed Gate Turnstile