Research On Vision-based Detection Technology Of Printing Defects

Printing defects usually appear on the surface of products which could bring huge economic losses to human beings.The traditional detection technology of printing defects is based on human eyes and the detection performance is not satisfied.Detection efficiency is low and there exist many wrong inspections.As the development of computer vision,related technicians start to combine vision and detection which becomes a research hotspot gradually.Therefore,this thesis focuses on the vision-based detection technology of printing defects and the main contributions are as follows.Firstly,a vision-based detection system of printing defects is designed.This system is composed of three modules which are image acquisition,image processing,input and output functions.According to the characteristics of printing detection,some reasonable requirments of functions and performances are formulated.The overall structure and workflow of the detection system are analysed.In addition,overall designs of three modules are analysed.Secondly,the image acquisition module,considering designs in terms of hardware and software,is proposed.Hardware design consists of analysing the functions of imaging system and motion system and making suitable lectotypes on the basis of characteristics of different parts.Software design includes studies on the control method of imaging system and motion system.Thirdly,the image processing module is analysed and put forward.A judgement criterion to printing images is set up.Those common printing defects are divided into three categories which are lump defect,line defect and point defect.Many kinds of algorithms such as image preprocessing,feature extraction,image registration,defect extraction,defect labeling and defect classification are studied.Besides,in order to improve the efficiency of defect labelling,a connected domain analysis algorithm based on improved run-length coding is put forward.Compared with traditional methods,this new method could highly increase the labelling efficiency.Finally,the detection system of printing defects is realized by combining the image acquisition module,the image processing module,the input and output module.System software and the functions of different components are introduced as well.The system is tested and the result shows that this detection system can timely and accurately identify printing defects.