Industrial Parts Shape And Size Of Image Detection Technology Research

Image measurement technology based on computer visions is a novel measurement method, which combines photoelectricity with digital image pressing technique. The paper uses image measurement technology for industrial components' shape dimension determination. The core technology, including sub-pixel detection, geometric element detection, error analysis and edge simulation image, are introduced in turn.A discrete orthogonal polynomial method is put forward for curve fitting sub-pixel edge detection algorithm, which decomposes two-dimensional curve fitting into one-dimensional curve fitting, and improves the reliability and computation speed of the algorithms.Errors resulting from discrete sampling of Zernike orthogonal (space) matrix is correct by error correction table. The edge detection precision is higher than 0.1 pixels though signal to noise ratio up to 20dB.By introduced the conception of orientated line for line detection, the precision and resolution of detection are improved dramatically. According to their shape characteristics, the parameter estimation algorithms of circles and ellipses are proposed based on the corresponding constraint condition. So the parameters are constrained and derived step by step. Then the reliability is improved while computation complexity is decreased.By combining Randomized Hough Transforms (RHT) and curve fitting method together, the geometric detection algorithms of lines, circles and ellipses are derived respectively and have advantages of automation, swift, precision and robustness as used in complex background.The approaches of linearity error and roundness error evaluating are presented, among which the least square method is emphasized. The conclusion that only when the little error assumption or little departure assumption are fulfilled the general least square algorithms are the suitable least square solution is provided. Based on the conclusion above, the least square solutions are derived using an iterative least square algorithm.The effects of point extended function, space discrete sampling, quantitative measurement, noise etc. on simulation image models are taken into account. A kind of high precision edge simulation image modeling process and the image modeling results are presented in the paper. The precision is less than 0.005 pixel.At last, using image measurement technology, the shape dimension detection system scheme of cigarettes is proposed. The simulation experiments results show that the following performances are reached. Detection speed is up to 60 pieces/s; length errors are less than 0.1mm; circumference errors are less than 0.05 mm; roundness errors are less than 0.05 mm. Themeasurement precision is higher than the standard of GB/T 5605-200*.