| With the increasing demand and output of mobile phone market,manufacturers put forward higher requirements for the accuracy and efficiency of mobile phone appearance inspection.The traditional manual inspection is inefficient.After a long time of inspection,the staff will have visual fatigue,resulting in false and missed product inspection,and there is a problem of mobile phone collision in the process of inspection.In order to solve the above problems and free the inspectors from the heavy repetitive work,the application of machine vision technology in the field of mobile phone appearance detection has become an important research topic.Based on the discussion of the project requirements,this paper designs a mobile phone appearance detection system based on machine vision,which includes motion control system,visual detection system and software operating system.According to the requirements of visual inspection,the visual system design and visual hardware selection are carried out.According to the characteristics of the target to be tested on each detection surface of the product,the product image is preprocessed,mainly including image filtering,visual positioning and geometric transformation.Aiming at the optimization of shape matching parameters,a shape matching parameter optimization method based on coordinate rotation is proposed.For character defect detection,the character defect detection method based on image difference is studied and analyzed.For character recognition,the character recognition methods of CNN convolutional neural network,SVM support vector machine and BP neural network are studied and analyzed respectively.Aiming at the defect detection of bar code,a detection algorithm based on the minimum circumscribed rectangle,the maximum inscribed rectangle and geometric features of connected domain is proposed.For color detection,a detection method based on RGB color component comparison and color similarity calculation is proposed.For parts missing detection,the detection method based on automatic threshold segmentation and geometric features of connected domain is studied and analyzed.Finally,the layout design of software interface is completed,and the detection effect of the system is experimentally analyzed.After verification and testing,the detection accuracy of the system reaches 98.652%,the continuous detection time of the system is 5.5 ～ 6S,and the detection efficiency is basically more than 600 per hour.All the detection data of the system meet the design requirements,and the detection cost and efficiency are much better than manual detection. |
