Automated visual inspection for the quality control of pad printing
Pad printing is used to decorate consumer goods largely because of its unique ability to apply graphics to doubly curved surfaces. The Intelpadrint project was conceived to develop a better understanding of the process and new printing pads, inks and printers. The thesis deals primarily with the research of a printer control system including machine vision. At present printing is manually controlled. Operator knowledge was gathered for use by an expert system to control the process. A novel local corner- matching algorithm was conceived to effect image segmentation, and neuro-fuzzy techniques were used to recognise patterns in printing errors. Non-linear Finite Element Analysis of the rubber printing-pad led to a method for pre-distorting artwork so that it would print undistorted on a curved product. A flexible, more automated printer was developed that achieves a higher printing rate. Ultraviolet-cured inks with improved printability were developed. The image normalisation/ error-signalling stage in inspection was proven in isolation, as was the pattern recognition system.