Shape-error Image Measurement For Shaft Parts

Shaft is one of the most important constitution in machine. Shape accuracy of shaft directly affects the using functions of machines. And because errors are inevitable, in part drawing in the form tolerance must be given to limit the shape error, the actual production process will inevitably have to measure the shape error. However, the traditional shape of the error of measurement accuracy often exist low complexity of the measurement process, or equipment is expensive, can not achieve non-contact measurement and other shortcomings, therefore, of rapid, accurate, non-contact measurement of shaft part shape error method in practice the production of great significance.In this paper, the computer vision technology and digital image processing theory, try a new digital image processing based on a round shaft, cylindrical shape and linearity error of measurement methods, design and build a shaft for the shape of the non-contact measurement system error, and the axis straightness, roundness and cylindrical were measured.The measuring system including digital image acquisition system, digital image processing system and output system of three parts, mainly from the light source, image sensor, optical lens, the image acquisition card, image storage, image processing software, computer and other components. The overall design of measurement system measurement program are as follows:1) The measured shaft fixed between the top two in the machine, the use of dividing heads turn every time a decile control axis angle, CCD camera in front of a circular shaft, around a circular axis equal portions of each corner then collected by a CCD camera image of a circular shaft, the shaft one revolution and, ultimately, a group of digital images. 2) The use of computers to filter the collected images after denoising, edge detection with subpixel edge technology and precise positioning technology edge information extraction part, on the edge of the information processing program to obtain the corresponding measurement data, and according to different measurement The mathematical model of the project, evaluation methods, the measurement data to be addressed, and ultimately be measured.Since any three-dimensional object obtained by the CCD camera after the two-dimensional images are formed by the combination of pixels to calculate three-dimensional space by two-dimensional geometry of objects, and then strike a shape error, it must be asked a two-dimensional pixel size and spatial correspondence between the geometric and spatial three-dimensional surface geometry of a point and its corresponding point in the image is the relationship between the geometric model of the camera imaging the decision, the geometric model parameters is the camera parameters . And due to the imaging condition, lens distortion and other factors, resulting in the same image in different locations, this correspondence is different, therefore, to accurately obtain this correspondence, in most conditions, must be experimental and calculation to get, this process is called camera calibration (or calibration). The size of calibration accuracy, a direct impact on the accuracy of vision measurement.For the part is the actual measurement, this paper used a simple, convenient and fast method for calibrating. Select the structure before the experiment, test pieces of the same size, higher precision axis as the "template" to use the camera when shooting fixed position, adjusting the light source and other experimental equipment, ensuring the same experimental conditions, the first card installed in the "template" axis capture an image, use the shaft diameter and the image of the correspondence between pixels, the partition coefficient calculated calibration. Then place the workpiece, and then shot under the same conditions, using the calibration coefficients calculated before, the programming can be calculated at the workpiece cross-section diameter of the corresponding value.Shape error of image measurement, in addition to the camera calibration, we should also use edge detection techniques and subpixel edge positioning technology accurately extract the image edge, and then programmed to calculate the edge information using the form error.Although the edge detection techniques have been enough to detect such a simple edge of the shaft, but the shaft is a metal parts, especially in machining, the surface is very smooth, so the reflective surface is severe, and as reflective and cause an axis Parts of the edge is no longer a typical step edge, while there will be significant warp phenomenon, which gives the edge detection shaft has brought some difficulties. When using the edge detection operator used to directly detect the edge of the binary image obtained in reflection at the edge will be some false, and the existence of these pseudo-edge programming to the post will bring great difficulties. This article address this issue the approach is: first with the edge detection operator Canny edge detection, and then get on the program a binary image processing, artificial edge are convinced that the area is definitely not the pixel value, thereby very simple and effective addition to the false edge.In order to improve the shape error of shaft measurement accuracy, the paper's edge in the use of pixel-level edge detection technique based on the extraction, but also by the sub-pixel edge detection of image edge subpixel edge positioning.The current sub-pixel edge detection algorithm based on moments, based on the least squares (fitting) and the method based on interpolation. Subpixel edge location technology used, can improve the accuracy of edge detection, but overall, the proposed sub-pixel edge legally obtained accuracy is usually higher than the other two methods, and the fitting is robust to noise and stability more. In this paper, a least squares fitting, and the characteristics according to the edge of the shaft, using the side of the warping of the edge of the step function instead of an ideal step function convolution, allowing more accurate subpixel edge location.In the image measurements, obtained an accurate edge information is critical, however, the measurement for different goals, establish accurate assessment of the model will directly determine the form error measurement accuracy.This roundness error in the measurement is used when the "Extremes." The specific method is: the use of CCD cameras in the pictures taken by sub-pixel edge detection and edge localization technology obtained the coordinates of the upper and lower edge of the value, calculated in the measured cross-section pictures the difference between the upper and lower edges, is the diameter value, Find the maximum diameter and minimum diameter, whichever is the difference of half a roundness error value. This evaluation method can be excluded due to the installation of the measured axis measurement errors caused by eccentricity, which do not have to use error separation technique for error separation method is more simple. However, in order to further improve the measurement accuracy, the paper also takes into account the shaft was to measure the impact of the results, and programming was removed.Side of the single axis does not reflect the image of the cylinder axis error, this article analyzes the use of multiple images feasibility of measurement error column, and the establishment of a cylinder using digital image measurement error of the mathematical model, based on the mathematical model , using the optimized algorithm for minimum zone evaluation of cylindrical, but also removed to measure the eccentricity of the impact of the results, and further improve the measurement accuracy. This experiment shows the balance between efficiency and accuracy of measurement of the number of cross sections and rotation options.Shaft line in the measurement error, the use of single side of the image axis is the measured degree of contour lines, and obtained by rotating all the pictures a week, you can find the axis straightness error. This article establishes the corresponding mathematical model of the algorithm for evaluating and measuring cylinder with the same degree of error of the method used excluded due to the bending axis to the impact of measurement results.Finally under different experiment condition, measurement and contrast analysis were done on several group of shafts which had different dimensions using the proposed measuring system and precise measure. The experiment shows that high precision and stable result can be obtained when measuring dimension of shaft making use of machine vision and digital image process technology. The experiment result proves the proposed method is effective and practical, and the method has merits as simpleness, convenience, fast calculating speed and high precision, which settles a base for real time and speediness detection of shafts.Along with the improvement of shoot precision of array CCD camera edge extracting technology, the mentioned method is attempt to be utilized to conduct non-contact measurement in real manufacturing process.