Innovative surface inspection with multispectral technology and artificial intelligence (INSPECT)

Important results from the project

The project will evaluate new technology for detecting and classifying surface defects online during production. The project will evaluate hardware and AI algorithms. Investigate the possibilities with an advanced vision system with a multispectral camera. Analyze the potential with AI algorithms. The project will deliver a plan for an implementation and implementation of a surface inspection system in an industrial environment. The project met all set objectives.

Expected long term effects

The results show that AI is a powerful tool for automatic defect classification. The industrial trial with the vision system and the multispectral camera worked in harsh environments and on hot material. The project helps to inform the steel industry about new opportunities for increased quality assurance of its manufacturing processes. As at present, large manual resources are required to subsequently locate, assess and remedy surface defects. An automated technology able to detect and classify surface defects in production would provide large cost and resource saving.

Approach and implementation

In the project, the companies have analyzed their needs and requirements on a surface inspection system. Based on the specification, suitable equipment has been evaluated. Databases with different types of defects have been used for performance testing of different AI networks. The results have been analyzed and discussed at continuous project meetings. The results have been presented at meetings with Jernkontoret TO44 and at a webinar arranged by the project. In addition to the project members, 25 people from the industry participated in the webinar.