An Investigation on a Low-cost Machine Vision Measuring System for Precision Improvement

In this paper, we describe the investigation on a machine vision size measuring system to improve its precision on the basis of the inexpensive devices from the viewpoint of industrial applications. The uniformity and stability of the system were analyzed. The results showed the maximum gray value standard deviation of the edge as 2.6 pixels, and the maximum error of edge detection results was approximately 9 pixels (0.279 mm). The traditional noise reduction algorithms were applied to reduce random noise and dark current noise, and a novel uniform-background algorithm was proposed to improve the uniformity of image background. In addition, a calibration method based on the average gray value of the specified areas was developed to correct gray value errors of the left and right edges. A large number of experiments were carried out using the combined methods, the results showed that the measuring speed was approximately 1 piece per second, and the maximum error of lengths measured by the proposed method was within 1 μm, whereas the maximum error of uncalibrated results was about 0.25 mm. The measuring precision and speed of the proposed methods can meet the requirement of industrial applications.