Novosibirsk Researchers Develop AI System to Check Assembled Orders
Researchers at Novosibirsk State Technical University have developed a prototype system that uses computer vision and neural networks to automatically check whether orders have been assembled correctly.
At small businesses, a single operator typically assembles and checks orders, either visually or by manually weighing the goods. With large volumes, concentration can quickly decline, allowing extra or missing items to slip into an order. Computer vision technology can recognize and count objects and check their dimensions in real time.
The university said there had previously been no universal off-the-shelf algorithm for checking assembled orders while accounting for the characteristics of individual parts, as most existing systems focus on detecting defects in products.
A software version is now ready for pilot operation, along with a prototype built around a webcam and ring light. The system measures product dimensions with an error margin of 1 millimeter. The researchers plan to reduce that margin by replacing the camera and eventually remove the operator from the process.
