Técnicas de visão computacional aplicadas na automação de processos metrológicos

Abstract

A common problem found in the calibration laboratories is the reliability of the results obtained in calibration of instruments, especially when they do not have a built-in communication interface. In this case, the spending time is increased significantly and the calibration can be subject to human error. In this context, many approaches based on computer vision have been proposed in the literature for automating calibration processes. Thus, four approaches of computer vision for application to the area of metrology were proposed in this work, three of them aiming at the calibration of instruments and one at the evaluation of quality injection mould surface. The first algorithm is based on normalized cross-correlation for reading digits of the display of a digital manometer; the second uses radial projections with Bresenham algorithm to determine the position of the pointer in analogical dial gages; the third makes a segmentation of the image to determine the holes size of screens grano-volumetric; and finally the fourth algorithm involves the automation of the evaluation of injection mould surface in which the proposed algorithm applied to images generated by means of laser device could help to detect problems such as over-polishing ("orange peel") or risks. Results showed that the proposed algorithms presented high accuracy and performance and can be used in real time systems applied to calibration processes.