Computer Vision System for Dimension Control in the Prefabrication of Concrete Panels

Abstract

This paper presents a computer vision system for dimensional quality control of concrete panel formwork. The methodology combines corner detection with photogrammetric principles to analyse overhead images of formwork assemblies, comparing detected geometries with design specifications. Validation on synthetic images demonstrates high accuracy, with mean absolute errors below 1.12 mm for dimensional and 0.02° for orientation measurements. While application to real-world factory conditions revealed challenges in corner detection requiring future improvements, the established framework provides a foundation for automated quality control in concrete prefabrication, enabling early detection of assembly errors before concrete placement.