A system for the quality control of ceramic products was developed. The system enables measuring geometrically complex products and detecting visual defects such as fragments, stick-ons, dirt and cracks with a minimal dimension of 0,01 mm × 0,2 mm and the cycle time under one second.
The system is composed of four independent sub-systems. Each sub-system contains a high resolution line scan camera and a special lighting developed by Wise Technologies. The processing is performed by a high performance quad-core computer.
Each sub-system inspects one surface of the product moving on the conveyor belt. The quality of the product is determined by summing up the results of all the sub-systems, which conduct digital image processing independently from each other. At the same time, the sub-systems are mutually synchronised, enabling constant conveyor belt speed and simplifying the manipulation.
Different challenges in the fields of machine vision and software architecture had to be addressed in the system planning and development phases.
A complex processing of four 25MB images in less than one second had to be ensured. The processing includes measuring 75 different measures in a complex product geometry, determining an adaptive region of interest in the image where the defects are detected, and the detection of the defects itself, which the existing single core processors are currently not able to handle. For that reason, we developed an algorithm for defect detection, where the processing is executed in four parallel threads running on four dedicated cores.