Research On Real-time Defect Detection Of Presswork System Based On Image Processing

With the development of science and technology, the degree of automation of themodern printing industry constantly improve, the traditional defect detection methodsin printing industry no longer able to meet the modern production requirements due totheir high cost, low efficiency and strong subjectivity. Therefore, research on theonline defect detection technology based on machine vision and digital imageprocessing has high practical value and important practical significance.After in-depth research on the relevant principle of online defect detection, thereis a more detailed discussion about several image processing technology that is veryimportant to this system. In this paper, put out a design scheme of the hardware andsoftware to the defect detection system, and implements the design is verified byexperiment. Around the basic flow and main technology of the online defect detection,this paper had done the following research:(1) This paper discussed the hardware platform of the online defect detectionsystem first, analyzed the functions and selection reference of main hardware devices;(2) Preliminary image processing can remove the noise or other unstable factorsthat affect the image, and guarantee the detection and recognition accurate andreliable. This paper introduced the two key preprocessing of presswork, that isimage denoising and image registration. At first, median filtering method is chosenfor image denoise in this paper. Depending on the type of primary noise thataffecting the quality of printing images, proposed the median filtering method shouldbe chosen in this paper. Then, Compared the two kind of image registration methodthat is based on gray information and feature extraction, and ultimately adopted SIFTimage features for image registration. Combining the advantages and disadvantagesof the SIFT algorithm and the characteristics of online detection system. It putforward to reduce the size of the match area to reduce the amount of calculationfeature points and guarantee the precise characteristics of these points. Make thealgorithm has better adaptability in the System in this paper.(3) It summarized several typical defect detection in image processing technology. Proposed image difference method based on wavelet packet transform.Using binarization and morphological methods to extract the defect area of thedifference image. Then use the Blob connectivity analysis to get the position andgeometric features of the defects, and marked the defects in the printed image.In this paper, MATLAB R2008a is used to verify the effectiveness of thealgorithm, and online defect detection system was realized in C#on Visual Studio2012development platform.