Measurement process control by means of algorithms based on digital image processing

The real-time verification of the correct functioning of a measurement system is not only seldom performed, but also done using fairly ineffective statistical tools such as traditional Shewhart control charts. By considering measurement results reported in a control chart as a digital image, two digital-image-processing algorithms are proposed for the early fault diagnosis (detection and identification) of measurement processes. The first algorithm exploits a uniform-approximation technique to face mainly problems of the measurement process and/or instrument drifts. The second is based on the Hough transform to approach cyclic phenomena also for low signal-to-noise ratios. Results of the numerical characterisation and experimental validation demonstrate the practical applicability and the effectiveness of the proposed algorithms