Design And Implementation Of Mobile Phone Case Defect Detection System Based On Machine Vision

Mobile phone is an important communication equipment(device)in modern society.As the most common mobile phone protection equipment,mobile phone cover can protect the mobile phone while there are higher and higher requirements for its appearance.At present,most of the defect detection for mobile phone covers in production still adopt manual detection methods that is difficult to achieve the standard of satisfactory detection quality.Besides,it requires high labor cost,which seriously reduces the detection efficiency.In recent years,with the rapid development of machine vision detection technology,people prefer to use automatic detection with computers instead of manual detection,which also provides a theoretical basis for the defect detection of mobile phone cover.The defect detection of mobile phone cases in this dissertation includes two subjects:color anomaly detection and surface defect detection.color anomaly detection aims to classify mobile phone cases of specified colors from other color mobile phone cases,and the surface defect detection is to classify the surface defects(scratches,missing blocks,white spots,lack of material)on the detection surface of the mobile phone cover and the non-defective mobile phone cover.Combined with machine vision detection technology,designs and achieves a detection system for color of device cover and surface defect.The main contributions are as follows.(1)Hardware architecture design of the detection system.By studying the optical characteristics of the mobile phone cover,the imaging module equipment and architecture is determined after theoretical analysis and image acquisition in the laboratory for many times.After that,by choosing the right frame,simulating the manufacturer's assembly line and selecting the motion control module,the imaging equipment and pipeline could be installed according to the recording architecture to complete the hardware design.(2)Software design and algorithm analysis of detection system.Based on the design scheme of the detection system,the software flow design and parameter settings are carried out.The main design procedures are: camera driver module,motion control module,detection algorithm module and result display module.After the image is collected by the detection system,the common image processing algorithm is used to color anomaly detection and surface defect detection of the mobile phone cover.For color detection,by calculating the average value of RGB in some areas of the surface comparing with the standard template,a reasonable color detection threshold is obtained,and the threshold is used for template matching and classification.For surface defect detection,different methods are analyzed.They are traditional template matching surface defect detection algorithm: end-to-end semi supervised image anomaly detection and segmentation framework based on deep learning,YOLOv3 algorithm and surface defect detection method based on deep learning segmentation.(3)Algorithm optimization and experimental evaluation.According to the comparison and analysis,the most suitable method and the way of improvement are found.In addition,by introducing residual networks,an improved defection method based on deep learning segmentation is implemented.On the other hand,the detection speed,detection accuracy,stability and accuracy meet the requirements of manufacturers,which means automatic detection of mobile phone cover can be achieved.Experimental results show that the proposed method is of great value.