Key Engineering Materials Online: 2014-08-11 ISSN: 1662-9795, Vol. 625, pp 305-309 doi:10.4028/www.scientific.net/kem.625.305 2015 Trans Tech Publications, Switzerland Automatic optical measurement of high density fiber connector Chang Chun-Li 1,a *, Chen Chun-Jen 2,b, and Wu Wen-Hong 1,c 1 Instrument Technology Research Center, Nation Applied Research Laboratories 20, R&D Rd. VI Hsinchu Science Park, Hsinchu, Taiwan 2 National Formosa University, No. 64, Wunhua Rd., Huwei Township, Yunlin County, Taiwan a hankchang@itrc.narl.org.twmail, b cjchen@nfu.edu.tw, c wenhong@itrc.narl.org.tw Keywords: Fiber connector; AOI; Subpixel; V groove; Collimated light Abstract. This paper will develop an automatic and fast optical measurement system for high density optical fiber connector measurement, which base on the optical project method. It will meet the lot of optical fiber connector measurement need. This system includes a high intensity and high collimated LED source, a high accuracy and high stiffness linear stage, 2 multi-degree-of-freedom adjustment mechanisms and a high resolution coaxial telecentric imaging system and image processing software. The system of measurement results and commercial measuring instruments comparing the results of its error is less than 0.6µm. To measure a V groove needs 750ms, this result may satisfy the requirements of automatic inspection. Introduction In recent year, the cost of optical fiber related product has significantly dropped; the optical fiber has been promoted into the terminal consumer market form the business market. In the future the optical fiber will become the computer standard communication interface and substitute the wire type transmission line. The optical fiber connector is the key component in the optical fiber device. In the early, the market acceptance of high density fiber connector is low because it is limited by the manufacture technology and market requirement. In recent years, the market acceptance of high density fiber connector becomes higher because the internet bandwidth requirement grows very fast. Because the quality of optical fiber connector will affect the optical fiber transfer performance, to ensure the optical fiber connector quality is very important. Although the existed scanning probe machine have high measurement resolution, but they don t suit the on-line and fast optical fiber connector measurement. The geometry of fiber connector and principal dimensions and tolerances are shown in figure 1. The tolerance of pitch of V groove is small range, to satisfy this condition moving direction must be within the scope of the position of tens of nanometer resolution capability. Traditionally used an optical interferometer position feedback devices, such devices are expensive and not conducive to apply to comprehensive inspection.[1]. Figure 1. Optical fiber connector and dimensions of connector. Because of the rapid growth of scientific technology, the production process has also come to accelerate the speed, precision products increasingly stringent requirements, the use of human vision as an inspection tool is inconsistent with the actual requirements, new inspection methods continue to 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-05/03/16,11:46:12)
306 Precision Engineering and Nanotechnology V be developing. Machine vision inspection system with quantitative characteristics and high repeatability satisfy the needs of modern production lines. In recent decades the rapid development of computer hardware and software technology, machine vision has been quite widely used in the automated measurement and detection of surface defects. Early computer vision measurement system does not to be applied to high-precision measurements due to the limited resolution of the sensor. However, based on sub-pixel algorithm can improve the resolution of computer vision measurement. The sub-pixel algorithm is the use of mathematical methods to the basic unit - pixels, which is divided into smaller parts to be able to improve the resolution of the measurement system. [2-3] Räsänen et al application sub-pixel technology to measure the thickness of flat glass [4], Lin et al are using sub-pixel algorithms to improve the correlation like method (CLM) resolution, and applied to measure the distance between the glasses of LCD panel [5] This paper will develop an automated optical inspection(aoi) system instantly detect V groove processing quality, the quality inspection system can instantly put the state back to the manufacturer so that the manufacturer can be tuning the production parameter in accordance with the quality of the finished product, the manufacturing system can achieve a stable manufacturing quality level. In this system, using high-resolution linear encoder as position feedback device, and then with sub-pixel methods to improve the resolution of the system, that does not require laser interferometer can meet the request of V groove measurement. Image Processing and Inspection Method Subpixel Algorithm Initially the one-dimensional (1D) sub-pixel edge detection studies, mostly using 1D ideal step edge mode because the real edge is continuous and susceptible to external interference. Using a 1D edge detection method for image subpixel edge estimation, the entire image to be scanned in order to be classified as units of processing, in order to obtain images of the subpixel offset edge, there are two major disadvantages, susceptible to noise interference, large amount of computation required. Two-dimensional (2D) subpixel edge detection method because only part of the analysis for edge. Compared with 1D subpixel method, 2D subpixel method is to save computing time. Therefore, this paper uses a two-dimensional subpixels using algorithms, the first convolution of the image with a Gaussian smoothing mask, followed estimation x and y-gradients using simple gradient operation G x = [ 1 0 1] and G y = [ 1 0 1] followed by non-maximal suppression to locate the local maxima. To do hysteresis thresholding on the gradient magnitude value on the ROI for V groove edge extraction, and finally on the edge of the area subpixel computing, subpixel interpolation by least-squares of a quadratic polynomial to the gradient normal to the detected edge[6], flow chart shown in Figure 2. Figure 2. Flow chart of subpixel edge estimation. Experimental Method The inspection system is constituted by a two-phase stepper motor, a stage with 1 mm lead screw with 50nm resolution linear encoder for position feedback, 10X CCTV lens and digital camera (1024 *1024 pixel, 1/2 " sensor) and a red LED collimated light, The schematic and configuration of inspection system are shown in Figure 3. In the geometry status between optical fiber connector and the camera is an importance factor of image quality. The inspection is performed before correction lens module and connector between
Key Engineering Materials Vol. 625 307 perpendicular. Figure 4 shows the images without correction and with correction. When the installation is complete, first of all, move the connector until the first V-groove in the image appear. User settings edge Region of Interest (ROI). ROIs are finished with the settings, the stage automatically in accordance with V groove of pitch movement, capture images and edge analysis to calculate the V groove of the pitch. Figure 5 shows the ROIs and inspection flow chart. Figure 3. a) Schematic of inspection system, b) The configuration of AOI system. a) b) Figure 4. The image shown in a) was grab without any adjustment of the camera. In the image can be observed the gray value is brightest at the bottom of the V groove, but the value is low in top of the image, and V-groove have a shadow edges. The image shown in b) was grab with well adjustment of the camera. In the background of the image the brightness uniformity is more than a). Experiment Result and Discussion Figure 6 shows fiber connector pitch measurement result and commercial measuring devices (KEYENCE VK-9700) measure the same sample result. The maximum error is 0.58µm appear in the 37th groove. Measuring time including the stage moving, image capture, the image processing, the average measuring time is 750ms.
308 Precision Engineering and Nanotechnology V Figure 5. a) ROI of edges and b) flow chart of inspection Conclusion Figure 6. AOI measurement result and measurement result of VK-9700 This paper presents a low-cost optical fiber connector automatic inspection system is constituted linear encoder-based motion control unit, a two-dimensional sub-pixel algorithms and image processing unit. To compare AOI measurement results and commercial measuring instruments measured results of the error is less than 0.6µm. Experimental results show that two-dimensional subpixel algorithms can be applied to measure the dimension of the transparent material. In the future, the related technology also can be used to other automatic optical inspection system for micro structure measurement application.
Key Engineering Materials Vol. 625 309 Acknowledge This work was supported in part by the National Science Council, TAIWAN, under Grants NSC 102-2622-E-492-010-CC3 and NSC 102-2622-E-492-013-CC3 References [1] Information on http://yagishitagiken.jp/original8.html#hikari [2] K. Jensen; D. Anastassiou, "Subpixel edge localization and the interpolation of still images, " Image Processing, IEEE Transactions on Vol. 4, no.3, pp.285-295. (1995) [3] E.P. Lyvers, O.R. Mitchell, M.L. Akey, A.P. Reeves, "Subpixel measurements using a moment-based edge operator," Pattern Analysis and Machine Intelligence, IEEE Transactions on Vol.11, no.12, pp.1293-1309. (1989) [4] J. Räsänen, K.E. Peiponen, On-Line Measurement of the Thickness and Optical Quality of Float Glass with a Sensor Based on a Diffractive Element, Applied Optics, Vol. 40, Issue 28, pp. 5034-5039. (2001) [5] C.S. Lin, J. Cho, Y.L. Lay, C.H. Lin, N. Lin, The auto-measurement of the gap of LCD glass plates using sub-pixel accuracy estimation, Vol. 117, Issue 8, pp. 349-354. (2006) [6] P. Rockett, The Accuracy of Sub-Pixel Localisation in the Canny Edge Detector, In Proc. of the British Machine Vision Conference (BMVC), pp. 392-401. (1999)
Precision Engineering and Nanotechnology V 10.4028/www.scientific.net/KEM.625 Automatic Optical Measurement of High Density Fiber Connector 10.4028/www.scientific.net/KEM.625.305