Study On Monocular Vision Measurement System For Tube Pan Of Nuclear Steam Generator

This thesis is based on the inspection of the tube pan of Qinsan II 600MW nuclear steam generator studied the monocular vision measurement technique.The most important values that need to be detected are the diameter of hole, hole distance and bridge between neighbor holes. In the past, various types of tools had been developed for measuring the tube pan, such as a micrometer calipers and industrial measuring systems. However, most of these methods require the tools to contact with the tube pan and have disadvantages such as huge labor hours, lower speed, contact distortion and etc. In this paper, a new non-contacted technique based on monocular vision is presented.Firstly, hardware of the measuring system is introduced in this paper. The designed computer vision system for tube pan inspection includes a calibrated camera, special scale bars, lamp-house and a computer with related software. The camera is used to take pictures; the scale bars are used as reference length for correcting image and getting scale factor and the lamp-house is used to illuminate the tube pan. By image processing in the computer, a 2D position of the holes on the surface of tube pan is created. The coordinates are stored in a database, which can be queried at any time for measurements via a graphical user interface. So, the parameters of the holes can be calculated.Secondly, principle of the computer vision measuring system is studied in detail. The image processing methods are studied to calculate the image coordinates of the center of holes, the first step is the accurate detection of holes'edges. Holes'edges are detected using Canny edge detector; edge linking; segmentation of the edges chain at points; and then using least-squares method to best fit the ellipse. The accuracy of the center coordinates of the hole can reach sub-pixel level; finally based on the length control the system gets the corresponding coordinates and judges each hole is right or not.Thirdly the main error sources of the system are discussed. Estimating the accuracy (or uncertainty) of measurement requires a careful treatment of the sources of error, including the error in measuring the image points, the error in estimating homogeneous matrix, the error in calibrating camera, the error coming from system hardware etc.. The uncertainty model are built and quantitative analysis of the uncertainty are given in the paper.Finally, the monocular vision system is produced in VC++6.0 and used to inspect the tube pan in product line. The inspection dimension of the system is 400mm×400 mm, with about 100 holes being inspected at a time in 10 minutes, and the accuracy is better than 0.05 mm, which proves the vision measuring system is meeting the requirement.