Research On Automatic Detection Method For Defects Of Mobile Phone Glass Cover

The progress of science and technology and the development of industrialization have enabled smartphones into thousands of households.People’s requirements for mobile phone are further improved,while mobile phone glass cover is the most direct part of smartphones and human interaction,and the quality requirements are also strict.The world’s daily demand for glass covers is several billion pieces and has increased by ten percentage points from ten percentage points.Huge market stocks and growth potential make the glass cover into the focus of many businesses.The quality test of the mobile phone glass cover is an essential procedure as the last level of production,but also to ensure the accuracy of the test,but also to ensure the detection speed.Most manufacturers still use the most primitive method of artificial testing,but manual detection has been obvious,which also provides development opportunities and prospects for the automation of the glass cover.This article studies the existing testing technology,improves the traditional glass defect image acquisition technology and image preoperative technology,and designs a new type of glass defective testing plan;combined with data set expansion related technologies improved deep learning glass defects plan.The main points of this study are as follows:Image acquisition of mobile phone glass defects is the first step in the research content of this paper.With the improvement of mobile phone glass cover technology,manufacturers have higher and higher requirements for the accuracy of glass defects.There are various types and sizes of glass defects,and various and dimensions of glass flaws are different,and various defects have a variety of ways to present on the glass cover,and different defects require different light source lightning angles to take clearly visible,plus ghost reflective Waiting for a multi-faceted reason,it has caused its image acquisition.This article combines existing image acquisition techniques to build a platform for mobile phone glass cover image acquisition system.Including the design of the light source,the selection of the camera,and the installation of the unit,and the design of the glass moving console is detected.Since obtaining high-quality images is still a difficult problem in the glass defect detection industry,it is particularly important how to preprocess the captured images to obtain images that meet the requirements of later defect identification and classification.Since the traditional median filtering method for image processing is a kind of full-image generalization filtering,part of the information of the image will inevitably be lost,the image will become blurred as a whole,and the edge information required for glass defect detection will also be lost.According to the characteristics of glass defect images,this paper improves the traditional median filtering algorithm,and adopts a new filtering algorithm with threshold judgment suitable for glass defect images to judge whether the target pixel should be filtered.It can not only enhance the image,but also minimize the loss of information.In contrast,the effect of image processing is more obvious,and the image quality is significantly improved,which lays the foundation for the subsequent identification and classification of defects.In this paper,in terms of judging the defects of defective images,according to the characteristics of mobile phone glass cover images,a detection method for mobile phone glass cover defects based on blob algorithm and template positioning is proposed.The method uses the template to position and straighten the obtained glass image,and then extracts the head and tail and the main screen in the middle,respectively,detects the head and tail and the main screen respectively,and demonstrates the feasibility of the method by experiments.In addition,a method of defect identification based on template matching and symmetrical aberration is proposed theoretically.The method is conceived based on the geometric characteristics of the cover image.The head and tail use traditional template matching to perform aberration operation.The middle main screen performs left and right differences according to symmetry,and uses a two-stage fusion method to detect screen defects.The birth of neural networks and deep learning and development provides methods for the test of glass defects.This paper studies the classification approach of current glass defects,and systematically studies the depth study of glass defects classification methods.However,the laboratory is currently limited,and the depth learning in this case may increase the likelihood of its model.In the case of the laboratory glass,the data set is expanded in this paper,the required data set is expanded to obtain a sufficient number of data sets,and classified according to the pre-formulated template.Training,get a network structure model that conforms to the category.Experiments show that the category model under this method is compared with the traditional method and the method without data set,its classification effect is significantly improved.In this paper,experiments are demonstrated in each chapter and the results are analyzed to verify the effectiveness of the method.