Algorithm Research And Implementation Of End Plug Surface Defect Detection Based On Machine Vision

With the increasing demand for new energy,nuclear energy,as a new type of energy,has been concerned by more and more people with its high energy concentration and rich energy storage,and nuclear power is an important form of nuclear energy use.Possessing the advantages of environmental protection and good efficiency,there is a higher demand about using nuclear power in the safety,especially the safety of nuclear reactor,and fuel end plug(hereinafter referred to as end plug)is an important component of nuclear reactors,so the defects on its surface affect the safety of nuclear reactors directly.At present,the detection of surface defects is still in the manual stage,and there have labor intensity and low efficiency problems.On account of these problems,it will be a better method to use machine vision and image processing technology.In this paper,based on the research of the defect features,two defect detection algorithms are designed for the two parts of the surface defects of end plug – the side surface defects and the center hole defects,as well as the simulation experiments of algorithms.In addition,a hardware test platform has been designed to detect defects on surface of end plug,and developed corresponding software.The main work of this paper lies in the following aspects:(1)Design and construction of hardware test platform for surface defect detection of end plugThe design of the hardware testplatform of “rotating platform + LED forward light source + line-scan digital camera” and “LED backlight + area-array camera” are used to acquire images of end plug's side surface and center hole,and take advantage of PC to deal with acquired images.(2)Research and design of surface defect detection algorithm for end plugThe detection process of “image preprocessing + defects extraction + defects classification” is used by analyzing the defect characteristics of side surface and center hole of the end plug.In the process of each step,the corresponding detection algorithms are designed according to the characteristics of the end plug side surface defect image and center hole image.(1)Detection algorithm of side surface's defectsThe images of side surface are acquiredby line-scan digital camera with forward light source.According to the defect characteristics of side surface,it adopts the preprocessing result throughmedianfilter and background compensation method based on polynomial.Next,it acquires the defects through threshold segmentation and removes non-defects with morphological analysis to improve the accuracy.Finally,it classifies defects of side surface through support vector machine(SVM)classification.(2)Detection algorithm of center hole defectsThe images of center hole are acquiredby area-array camera with backlight.In this paper,the center hole defect is detected by polar coordinate processing of its edge ingformation.Specific process: Firstly,it adopts the preprocessing result based on Gray gradient filter method.Next,it conducts the threshold segmentation with advanced two-peak method,extracts the edge information through edge detection,and acquires defect information through polar coordinates.In the end,it classifies the defects through the binary tree classifier.(3)Design and implementation of surface defect detection softwareIn the VS2010 software development environment,based on the C ++ language,the plug surface defect detection software is designed and developed by using MFC framework,multithreading technology,Open CV visual development library and other programming technology.The software's function that acquisition of side surface image and center hole image of end plug can be completed by interface operation,and it also can detect defects and save defect information,etc.In this paper,the research and design of the surface defect detection algorithm for end plug and its software implementation,through the machine vision technology to complete the defect detection,reduce the labor intensity and improve the efficiency of end plug detection.The results show that the detection accuracy of the side surface defects can reach more than 86%,and detection accuracy of the center hole defects can reach 90%,which basically fulfill the surface defect detection requirements of end plug.