The Design And Implementation Of Visual Inspection System For Lamp Glass Shell

Light bulb by firing,racking,sealing,filling head,inspection and other aspects,in order to achieve high-speed automatic production.In the lamp glass shell molding process,due to the complexity and the high speed of the production process,the low precision of the mechanical control system,there are inevitably exist a variety of defects in the manufacturing and assembly and other aspects,such as appearance defects including broken glass,dirt pollution,surface cracks and dimensions defect including length,diameter,thickness,which can not enter the next bulb lights manufacturing sectors,must be identified and sorted.The ’traditional manual inspection could not meet the requirements of product quality,safety and high-speed,large-scale automated production lines,mainly displays in the low artificial detection precision,strong subjectivity,poor repeatability,inconsistent test results,the high unchecked and false detection rate,which is difficult to meet the demand of high-speed production line.Therefore,it is of great economic and social significance to study the automatic detection technology to replace manual testing.This paper introduces the relevant key technology of auto lamp glass shell visual inspection,designed the mechanical structure and electric control system.In the image detection processing,studied in depth of the various visual detection method,a prototype is developed,and a software system of image processing algorithm is developed,and the whole system is applied to the automatic production line.The research work of this paper,main achievements and innovation points include the following aspects.1.This paper describes the automobile lamp bulb detection situation,on the basis of summarizing the research results at home and abroad,combined with the actual demand,we put forward the overall scheme of visual detection system.In the mechanical structure,we design the chain wheel transmission structure,through acquiring a sequence of images to achieve a full range of glass detection.The electrical control system based on PC and PLC is introduced in detail,based on analyzing the optical characteristics of glass shell we designed a set of optical illumination solutions.2.In the image preprocessing algorithm,we use gray transform to enhance defect characteristics.In the image denoising,comparing the several grayscale morphology and image filtering method for noise reduction effect.And put forward a method for detecting target location based on edge points.In the glass shell edge feature segmentation test the traditional edge detection algorithm.In the geometric primitive detection,the traditional Hough conversion efficiency is not high,Studies repeatedly iterative fitting algorithms in detecting lines and circles application.On the basis of the above,we made a detailed introduction for the size measurement of the glass shell.3.Designed the glass shell defect body database,using the BP neural network algorithm for defect classification.In glass shell neck crack detection,this paper proposesd a template matching algorithm,the experimental results show the algorithm error rate is high,based on this method we puts forward a method called vertical integral projection to detect cracks,and applied feedforward neural network to detect the crack,the experimental results show that these two algorithms are feasible and effective.4.We developed a prototype and a software system used for industrial management and monitoring,designed a set of database for saving images and data,it provides a good experimental platform for expanding theory in research and more practical application occasion.After industrial application,the detection speed can reach 5500/per hour,and the detection accuracy can reach 0.01mm.The practice proves the practicality of the prototype and software system,the validity of image methods.It basically satisfied the automatic production line in real-time,high-speed and high-precision detection requirements.