How is the 24-hour unmanned convenience store technically realized?

How is the 24-hour unmanned convenience store technically realized? The current technology to realize the technology of 24-hour unmanned convenience store To do this, we must first solve the problem of identity recognition. Biometric identification technology is used: according to the structure of each customer's palm capillaries, a string of encrypted characters is generated, and this character is the user's "lifetime ID".

Mr. Chen Haibo pointed out that technologies such as fingerprint recognition and face recognition use external biometrics, which are easy to obtain and have low security. The subcutaneous capillaries are internal biological features and are more private.

In addition, the false recognition rate of subcutaneous capillary identification is also lower than other biometrics, only 3 parts per billion. This is because subcutaneous blood vessels have more abundant biometric points—about 500-600, while fingerprints have only a dozen or so.

In unmanned convenience stores, in addition to accurate identification, consumers must be continuously tracked and their behaviors analyzed. The company uses convolutional neural networks and machine vision to analyze the posture of consumers. Mr. Chen Haibo introduced that Amazon Go uses face recognition in this link, but the face is easily blocked. When consumers bend over, the camera may not be able to capture the face, so it is easy to lose. Body recognition does not have this problem, and the accuracy is also guaranteed, because the body shapes of two people cannot be completely integrated.

In addition to identifying consumers, product identification is also a major difficulty. The difficulty in recognizing goods with machine vision is to remove the background. For machine vision, whether the goods or the background are pixels, they need to be recognized. If an item is placed on a conveyor belt and a stationary tabletop, identification is relatively easy. But placed in the store, customers walking around, constantly moving products, and the light in the store is constantly changing, all of which will affect the accuracy of product identification.

To this end, current technology has developed a set of algorithms for real-time background removal - except for consumers and goods holding goods, the background is blurred. If you go into the store with a friend to buy, you hand over the product in your hand to your friend, and at the moment when the product is handed over, you will become the background blurred, so as to improve the accuracy of product identification.

The company's 24-hour unmanned convenience store has a special settlement area. Once the user enters the settlement area, it is assumed that the customer has decided to purchase the product, and the system will automatically initiate a deduction. As for how the user pays, it depends on the payment account bound when the user first registered. It is reported that Alipay, Ant Huabei, UnionPay, WeChat and other payment methods are currently supported.