Model-based block diagram recognition for model visualization verification - extended version

Abstract In safety-critical applications, domain-specific modeling remains insufficiently automated, requiring significant effort for qualification. One critical issue is ensuring correct visualization without relying on manual verification processes. In other words: “is what you see really what you get?” This paper continues an approach for the automated verification of visualizations, specifically targeting block diagrams, which are a common representation in domain-specific modeling. Our approach utilizes image processing techniques to recognize block diagrams and compare them with the original models to detect potential deviations. Relying on a qualified model database implementation, we can purely focus on the verification of the visual chain. We demonstrate the effectiveness of our proof of concept through a use case involving two graphical domain-specific languages. Realistic scenarios like intersections, rotated text, and a variety of overlapping vertices of differing sizes and shapes are included. The implementation successfully detects visualization smells in 16 unique test cases, highlighting its potential for improving reliability and reducing manual verification efforts in safety-critical domain-specific modeling environments. The presented software architecture is suitable to be a qualified visual checker tool in the aviation domain.