KamAZ has introduced computer vision for color control

2024: Completion of pilot project on implementation of computer vision system

On January 22, 2024, the company KAMAZ Digital announced the completion of a pilot project to introduce a system computer vision for quality control of the application of paints and varnishes of body parts in production. In cars KAMAZ 2024 computer , the vision system will be scaled on the paint line of the No2 Press and Frame Plant of PJSC KAMAZ.

According to the company, in the pilot project, the training of the computer model was carried out on a limited number of parts, as a result, the functionality for calculating the number of parts works at the level of 90-95%, and the determination of the quality of the primer of the product reaches 87-94%. The goal of the project is to reduce defects in the priming of parts by reducing the influence of the human factor in the detection of soil defects.

The main objective of the pilot project was to minimize errors in assessing the quality of priming parts before painting and reduce unforeseen costs. Due to the high impact of the human factor, the company incurred additional costs associated with the elimination of defects.

After the analysis of the problem of the appearance of a defect in the paint coating, it became clear that conventional methods of periodic control are not enough. Then a solution appeared that allows you to control the defects of the primer before the painting operation. The developed computer vision system determines the type of parts, their number and the presence of defects. Keeping the archive of the record and informing the responsible persons, it allows you to respond to incidents in a timely manner and prevent the transfer of defective products further through the process or ship defective products.

The solution consists of software and hardware. The software is based on computer vision algorithms that recognize the image of the product identifies their type, count the products on the tooling, recognize the quality of the soil, etc.

Also, the software part includes a simple web-interface for employees of the technical control department and a system for storing archive data (from the event log to the archive of video recordings).

The hardware consists of infrastructure elements (communication lines and communication equipment) and actuating elements (video cameras and server equipment).

We have been looking for a solution to exclude or reduce the "human factor" in quality control for a long time, but these decisions were mainly related to increasing the staff of controllers. The implementation was carried out turnkey in 9 months. The solution was implemented for January 2024 on one of the seven lines. In 2024, it is planned to carry out work to expand the functionality of the system and expand the coverage area to other production lines. told Vitaly Postnikov, Head of the Department of Technical Control of Press and Frame Production of KAMAZ PJSC

The production process in KAMAZ is complex and often the use of some standard solution is impossible in principle. Therefore, on this project, the KAMAZ Digital team worked in the format. R&D The solution grew out of the hypothesis that using computer vision, you can solve the problem with defects. We conducted a number of experiments tests and made sure that the solution was correct. Computer vision as a technology can be applied in almost all areas of production, but, of course, careful research and experience with such technologies is required. told Pavel Besedin, CTO of KAMAZ Digital