The first is visual recognition. For example, when your hand reaches into the shelf, color recognition is used to determine whether there are any other colors around your hand besides the skin color to determine whether you have any goods in your hand.

If there is no item when your hand reaches into the shelf, and there is an item when you leave, then it is considered that you have picked up the item, otherwise it is put back.

Another method is to identify the product area. For example, there was a product in this area, and then it is recognized that there is no product here, then it means that it has been purchased, otherwise it is put back.

The second is the sensor. For example, the gravity sensor identifies whether the weight of the shelf has increased or not, and identifies whether the product has been taken away or put back;

The infrared sensor identifies whether the light near the product is blocked and identifies whether the product has been taken away or put back.

Amazon calls it multi-sensor fusion technology, which is to obtain data from different sensors to increase the credibility and accuracy of judgment.