Image-based cast part recognition using intrinsic texture

Purpose and goal

The basic idea is to use the natural texture of casting surfaces for individual marking. The aim is to test this idea and develop a proof of concept as a basis for a full-scale project. The study therefore has no measurable objectives for a finished solution, but if the concept proves viable, effects such as reduced scrap, increased circularity and improved resource utilization can be quantified. A key objective for the project is the marking should be identifiable with at least 98% accuracy.

Expected effects and result

The project will verify whether the natural surface variation of castings can be used for individual marking. It aims to develop methods for creating and matching unique “fingerprints” and demonstrate high-accuracy identification in a relevant environment. The study will define requirements for surface properties, lighting, and camera technology and identify issues to address. If successful, benefits include reduced scrap, increased traceability, improved circularity, and efficient resource use.

Planned approach and implementation

The project is divided into four work packages: data collection and feasibility study, AI model development for “fingerprinting,” integration and pilot demonstration, and project management. Work includes image collection in lab and industrial settings, development and training of matching algorithms, testing robustness and performance, and creating a roadmap for industrial implementation. A final report will be published, and if results are positive, a full-scale project proposal prepared.