Research on the size of mechanical parts based on image recognition

Abstract Because the image measurement technology based on machine vision has the advantages of high accuracy, high efficiency and non-contact measurement, this kind of measurement technology has gradually become the focus of attention in industrial production measurement and detection. Based on the analysis of image measurement technology, this paper studies the measurement method of mechanical parts size based on image recognition and improves related algorithms. Specific research work is as follows: Design a measurement method of machine parts size based on image recognition, study and analyze the formation of noise, types and corresponding denoising technology, select a fast median filtering algorithm to achieve filtering. Polynomial interpolation is applied to the sub-pixel edge location method to extract the edges accurately. Some classical operators are studied and analyzed with the specific part image to be tested as the experimental object. Several classical operators are compared and analyzed through many experiments. Experiments show that the improved morphological gradient operator can effectively refine the image edge. The experimental scheme proposed in this paper can better realize the measurement of mechanical parts size, and the improved algorithm has significantly improved the accuracy than before.