Design Of Logo Defect Detection Software System Based On Image Processing Technology

Logo is an important symbol of the corporate image,so the correct and standard use of Logo is particularly important.With the rapid development of information technology,the amount of office documents has been increasing enormously.Meanwhile,the demand for Logo defect detection in massive office documents is also growing under this situation.However,the traditional manual detection consumes a lot of human and material resources.Besides,manual detection may lead to the wrong detection and missed detection of Logo defects due to fatigue loss.Consequently,the automatic Logo defect detection shows a widely applied future.In this thesis,taking the Logo of SB C&S company which is a laboratory cooperation as the main research object,a Logo defect detection method based on image processing is designed to solve the problems of Logo color,shape and blur.Firstly,this thesis designs a logo color defect detection algorithm,which detects the color defect of logo image by adaptive foreground and background separation and foreground color defect detection.The experimental accuracy of this method reaches97.0%.Secondly,this thesis designs a Logo shape defect detection algorithm,which detects Logo shape defects by training convolution auto-encoder to reconstruct Logo image and image difference,and reduces the image registration error caused by traditional image difference algorithm.The test accuracy reaches 97.4%.Thirdly,this thesis designs a Logo blur defect detection algorithm,which can effectively detect the Logo blur defect by calculating the sawtooth number of the image edge with full reference and the gradient of the image edge without reference,and the test accuracy reaches 96.6%.In addition,in order to verify the generalization performance of this algorithm,the generalization analysis of horizontal character Logo,such as Google,Nokia,etc.,shows that this algorithm is feasible in defect detection of horizontal character Logo.Based on the Logo defect detection algorithm,Logo defect detection software system is further developed in this thesis,which includes file management module,Logo positioning module,Logo defect detection algorithm module,asynchronous multi-task execution module and database module.The system adopts B/S architecture to support multi-users.Users only need to upload files,and the system will deal with defect detection tasks asynchronously,and finally return defect detection results.After experimental test,the system realizes the required functions.The detection accuracy of the system is 97.8% for whether the logo has defects,and 96.7% for the types of logo defects,which meet the requirements of the system design.