Design And Implementation Of Inspection System For Surface

The subject comes from the actual needs of glass manufacturers, the system of surface inspection of glass lampshade mainly uses an industrial cameras to get the image of glass products. And then, with the pre-processing and post-processing on the collected image, we finally detect the eligibility of the quality of glass surface. We permit the qualified products passing the pipeline properly and abandoned the substandard products.Currently, the glass shade is generally manual checking, mainly by the experience of workers,the detection process takes time and effort. Therefore, we design this system in order to improve the stability of detection, reduce the costs of personnel and material.The main contents are as follows:(1) Build the hardware system. In this paper, the hardware systems including PC,frame grabbers, industrial cameras, lenses, lighting, waste discharge devices and so on.This topic will be studied to select several different cameras, light sources. Other hardware requirements are also in accordance with the parameters of the actual system.(2)Research on the image algorithms. Digital image processing, is applied to one or some computing and processing with the digital computers and other related digital technology to achieve a desired purpose. In accordance with the running order, related to the image processing method described herein include: image preprocessing, image segmentation and image extraction. This paper analyzes several common pre-processing and classical algorithms of image analysis according to the characteristics of image of glass lampshade and proposed a clustering subtraction method for image segmentation and extraction of defect. We devised an suitable algorithm for the system on this basis, it is one of the hardest difficulty of this study.(3)The testing of the surface of glass shade. This piece consists of two parts,defects and hole sizes. Which the surface defects mainly refers to the bubble, cracks and the lack of the edge of the lampshade. This thesis, in each case, will propose the suitable algorithm of defect classification and circle fitting. Eventually, according to the results of the experiment, we select the least squares to detect the hole size and propose an improved BP neural network algorithm on the defect classification. This is anotherimportant difficulties of the system.In short, this glass shade detection system is stable, reliable, fast, high precision,real good. It can produce the false positives of manual testing to a minimum, meet the requirements of FAR of manufacturers. It has great economic value to glass manufacturers, while providing theoretical and practical reference for other related industries.