Applied Mechanics and Materials Online: 2010-11-11 ISSN: 1662-7482, Vols. 37-38, pp 513-516 doi:10.4028/www.scientific.net/amm.37-38.513 2010 Trans Tech Publications, Switzerland Image Measurement of Roller Chain Board Based on CCD Qingmin Liu 1,a, Zhikui Liu 1,b, Qionghong Lei 2,c and Kui Zhang 1,d 1 Hangzhou Dianzi University, Hangzhou 310018, P.R. China 2 Zhejiang Institute of Communications, Hangzhou 311112, P.R. China a qingmin61@163.com, b jizhi19@163.com, C leiqionghong@163.com, d zk7156@163.com Keywords: Roller chain, CCD, Roundness error, Image measurement Abstract. This paper proposes a new kind of detection method based on CCD to measure geometry sizes and roundness error of chain board of roller chain. Image is collected, and is input into computer, then is processed by a series of steps such as grey level transformation, image smoothness, lower filtration, and threshold value selection, etc. The measurement of the geometric parameters and the roundness error relies on the image contour. Roundness error is evaluated by Minimum Region Law. The diameter of the holes and the distance between the axes of two holes are calculated and measured by self-developed program. Roundness error of chain board is less than 8μm, and the distance error between the axes of two holes is 9μm. The diameters of the holes are 4.467mm and 4.461mm respectively. The reasons of measurement error are also analyzed. Theory analyses and experiment show that it is feasible to measure the geometric sizes and roundness error of roller chain by this method, which is both efficient and practicable. Introduction Chain transmission is an important machine transmission [1], its application is wide. When the inner and outer chain board relatively deflect, the sleeve can rotate freely around the pin shaft. To ensure fit quality, the size of pin hole in outer chain board, sleeve hole in inner chain board need strictly controlling, it s easy to measure and manufacture the external diameter of pin and sleeve, but there are more difficulties to manufacture and measure the internal size of chain board. The hole diameter, roundness of holes and center distance between holes have direct effect on the fit quality and chain pitch, so it s necessary to detect them quickly and precisely. The roundness error cannot be measured on roundness instrument because of small size of chain board, the size measurement precision cannot be guaranteed at a high level under the measurement of traditional multi-points contact measurement. Used CCD as image sensor, and measured those parameters mentioned above by image detection techniques. The measurement technology for chain board of roller chain has practical significance to improving product quality, enhancing company competitiveness. Image Processing Image Collection & Processing. A102 CCD digital camera, produced by Basler in German, is used to collect digital image, it s pixel number are 1932 1040, image area are 8.978 6.708mm 2. The digital images are transmitted into computer though IEEE1394 digital interface card. Fig. 1 a) shows original gray image of the roller chain board, it is the BMP file with 256 gray levels. a) Original image b) Noise-reduced image Fig. 1 Image collecting and pre-processing All rights reserved. No part of contents of this paper may be reproduced or transmitted in any form or by any means without the written permission of Trans Tech Publications, www.ttp.net. (ID: 130.203.136.75, Pennsylvania State University, University Park, USA-05/03/16,18:57:02)
514 Advances in Engineering Design and Optimization It is inevitable that the image will be polluted by noise during process of collection, transmission and restoration, in that case, image will become vague so that it is difficult to recognize edge and catch feature. Noise reduction is an important link of image processing, and smooth processing is used as preprocessing to reduce image noise. Mean filter s smoothing function will make image edge vague and have great impact on size and roundness error. Image collected by this system have uniform and single brightness of object and background, they have great contrast, other lines and details in image is easy to recognize, better result can be gotten by median filter. Fig. 1 b) shows the preprocessing image. Image Segmentation. The images are relatively simple in figure 1, binary converting in image processing can be performed by adopting gray histogram. Let gray range of image f be[a,b], binary threshold T, binary converting in image processing can be shown as follows: f T 1 f ( x, y) T ( x, y) 0 f ( x, y) T (1) Where, f T is binary image. Threshold is transformed from input image f to output image f T, if the pixel belong to object then f T (i,j)=1, otherwise, f T (i,j)=0. It is easy to segment image by threshold, the contrast of image in Fig. 1 is obvious, object is very dark, and background is very light, the value of threshold is set 130. Fig. 2 shows the result of segmentation, a) is binary image. The sketch figure with width of single pixel can be obtained through contour extraction which will be tracked and the refining treatment will be carried out on at the same time, As in Fig. 2 b). Measurement a) Binary image b) Contour extraction Fig. 2 Image segmentation System Structure and Measuring Principle. Measurement system consists of lighting system, CCD camera, IEEE1394 digital interface card, computer and corresponding software. The parallel light emitted by lighting system generates the shadow contour of measured object, then the shadow contour will be focused through lens system and forms image on CCD, the collected image will be transmitted into computer memory and then to be processed and calculated by software. In order to ensure the fit quality of components, each size must be controlled strictly. The roundness, diameter of pin hole and center distance between the axes of two holes are detected according to the results of image processing. Measurement of Roundness Error. Minimum zone method [2-3] is used to judge roundness error. It means that the area formed by two concentric circles which contain actual contour is minimum. The two holes in Fig.2 b) are factors to be measured. Let coordination of circle center of concentric circle be O(x 0, y 0 ), and the coordination of actual contour be x i, y i (i=1, n), x min x i x max, y min y i y max, then the distance between points and circle center are: 2 2 R x x y y (2) i i 0 i 0 Where, R min R i R max. Optimization model of discrete points is established as:
Applied Mechanics and Materials Vols. 37-38 515 min f X Rman Rmin (3) s. t. x0 xmin, x0 xmax, y0 ymin, y0 ymax, x0 I i 1,2,, n, y0 I i 1,2,, n Solve design variables X=[x 0, y 0 ] T, to make the radius minimum. In this case, use discrete penalty function method so that the discrete variable inequality constrained optimization problem can be converted into unconstrained optimization problem: 2n k k u X k i i (4) u1 ii X r f X r g r p p min, 1/ 4 1 Choose proper initial value to be iterated and get X=[x 0, y 0 ] T, which meet the iterating precision, that is the circle center coordination of minimum zone. Conditions for the termination of iterations is: K1 K f X f X (5) Use measurement software to measure Fig.2 b) according to the way mentioned above, iterating precision is 0.002, the datum are listed in Table 1. Table 1 Measurement results of roundness error holes X 0 (mm) Y 0 (mm) ƒ (μm) D 1 123.847 86.556 7.961 D 2 136.483 87.913 7.574 Measurement of Geometry Sizes. After image processing, the obtained pixels coordination of surface profile of chain board are discrete, Fig. 2 b) include curves and arcs. According to the pixel feature, image information have been converted into graphic information so that the calculation speed and measurement precision can be improved. Fig. 3 shows the design drawing of roller chain board, the sizes which are measured include pin pore diameter D and center distance P between two axes of the holes. The pin pore diameters are selected in sequence according to the extracted contour. On the basis of minimum zone method in roundness measurement, take the average value of minimum zone as the measuring result, namely D=(D MAX +D MIN )/2.The center distance P of pore diameters is represented by the center distance of minimum zone circle, after the experiment analysis and calculation, D 1 =4.467mm, D 2 =4.461mm, P=12.709mm. The pin diameter D and pitch P of 08A model roller chain are respectively 4.45mm and 12.70mm according to the manual. The results of actual measurement showed in Tab 1 are very close and the error is small. Fig. 3 External chain board drawing After image processing, the obtained contour coordinate is discrete pixel value, in order to calculate the actual size from object image pixels, the system should be calibrated. The proportion relation of sizes among object images are established through measurement experiment by standard
516 Advances in Engineering Design and Optimization gauge block. After determining the working distance of image measurement system, an image of standard gauge block M with known sizes should be formed in CCD, then the number N of pixels occupied by that object in CCD can be obtained, so that the calibration coefficient k=w/n of each pixel can be obtained, that represent the corresponding actual size of pixel. The actual sizes of measured object can be obtained when the measured object is put on this location. Some contrasting experiments showed increasing the resolution of CCD, reducing the lens distortion, using sub-pixel algorithm can improve the measurement accuracy. Error Analysis Error of vision system mainly consists of lens aberrations, light-sensitive pixel arrangement error and perspective error. The error of geometric distortion of imaging system is typical system error, it can be reduced by improving system resolution and adopting. The number of pixels of CCD selected by this system are 1392 1040, the size of chain board is less than or equal to 25 13mm 2, the maximum of corresponding chain board area of each pixel is no more than 2.5 10-4 mm 2. It is necessary to perform amplifying processing for improving measurement precision when the image is being collected. The formation of noises is due to the various interference during the process of imaging, digitalisation and transmission of images. Noises make pixel gray value of the images can not accurately reflect light intensity values of the point, so the image s quality will decline. Noises are mainly generated by the camera. In addition, pixel jitter at the time of the video and image are captured, is also an important source of noises. Using related filtering method can weaken and inhibit a variety of noises. Error of software algorithms mainly consists of binary image threshold selection error and the difference approximation derivative operator error. The process of calibration will introduce errors too[4]. In order to eliminate the system error, the second calibration method is used to determine the proportional coefficient k. Experiments have proved that the actual size of the measured objects and the corresponding number of pixels satisfy the relationships: W = kn + b. Where b is the system error of measurement, the values of k and b can be determined through the two calibration. That will eliminate the influence of system error on measurement precision. Conclusion Image measurement for the roller chain board is performed by using area array CCD, 1394 interface card and computer. Roundness error measured from the experiment is less than 8μm, pore and hole center distance error is less than 0.017, 0.011 mm and 0.009mm respectively. The lens aberrations, light-sensitive pixel array error and perspective error is the main source of error by analysing the reason of measurement error. The method which is described is particularly suitable for detecting small size, thin-walled parts, fragile pieces of soft and flexible parts, which can not be measured by contact method. Size error and roundness can be measured simultaneously in the same system. Find a new way for the error analysis of machine vision detection and relevant principles. References [1] Z.F. Zheng: Design and Application Manual of Chain Transmission (China Mechanical Industry Press, China 1992). [2] Y.P. Chen, X.B. Yue and X.Y. Zeng: Tool Technology, (2003) No. 4, pp. 39-42. [3] China Standards Press: The Standard Manual of Parameter Tolerance (China Standards Press, China 1995). [4] Z.Q. Qiu and Q.F. Yu: Journal of Engineering Graphics, (2001) Supplement, pp. 1-5.
Advances in Engineering Design and Optimization 10.4028/www.scientific.net/AMM.37-38 Image Measurement of Roller Chain Board Based on CCD 10.4028/www.scientific.net/AMM.37-38.513