Rapid vision-based dimensional precision inspection of mesoscale artefacts

This paper describes the investigation and implementation of image processing techniques converted into precision measurement techniques for feature inspection of mesoscale artefacts. The method resulted in a rapid, low-cost, and precise inspection technique. Better performances are obtained with the combination of the pixel uncertainty edge detector and the subpixel uncertainty detector along with camera calibration. This procedure revealed that the subpixel moment-based edge detector presents smooth edge boundaries that are closer to the true contour of the specimen compared to edges detected by Canny edge detector. In addition, it was observed that this smoothness is closely related to the mask size of the moments' calculation. The considered sources of error are the quality of the reference target and the accuracy of the camera calibration. A comparison with the results of the other instruments reveals that the uncertainty of the proposed inspection system could reach more than 5 μm with a much lower resolution and an absolute average difference of about 6 μm between measurements.