A prototype of a domestic machine vision system for comprehensive monitoring of production processes has been created at Novosibirsk State Technical University (NSTU). This was reported in the press service of the university.
The development automatically detects defects on the production line, monitors safety, improves work efficiency, and prevents accidents at enterprises.
The prototype showed an accuracy of 94.87%.
The system can be combined with the existing infrastructure of the enterprise without stopping production. The solution includes a device for improving network quality and a camera for detecting defects with an AI module. A data processing center is provided for system management.
The camera tracks each object, and then the video stream is analyzed using two neural networks. Defects and deviations from the norm are recorded using heat maps. All defects are classified into the categories "normal", "dent", "crack".
Currently, the university developers are creating a platform for centralized management of the entire system.
Earlier, www1.ru reported that a unique robotic dispenser for microelectronics and mechanical engineering was created by scientists at SPbPU.
Read more materials on the topic:
- Russian scientists have created a universal machine vision system for object recognition
- Not for color blind people, but for the sake of autopilot: in Kazan, a neural network was trained to recognize the colors of traffic light signals with an accuracy of over 90%
- Machine vision technology implemented at the Naberezhnye Chelny Cardboard and Paper Mill
