A study for a high performance parallel system for on-line defect detection in textiles

The work describes a preliminary study for the design of a parallel real-time system for the quality control of textile products. The foreseen structure, linear multi-spectral optoelectronics transducers and special light sources are used to scan both sides of the fabric; image data are send to a parallel real-time processing system, that should be able to detect defects on the fabric, which is moving at 1-2 metres per second, to determine and to mark the defect positions on the fabric. The system should also be able to define morphological and radiometric characteristics of detected flaws and to interact with the production process. The advantages produced by the automation of a process today performed in a manual way only are many: o an increase in the production through the higher inspection speed of the automatic process against the visual one; o an increase in the production due to the elimination of machine stoppages when the inspector rests; o an improvement in the process output quality through the substitution of the human criteria with constant mathematical criteria and through the elimination of the unavoidable moments of psychophysical stress of the operator; o a reduction and in prospective the elimination of the claims between the manufacturer and the customers due to the insufficient quality of the inspection.