Research And Application On Real-time Quality Detection Of Presswork System Based On Moving Target Detection

With the development of science and technology, The theory of machine vision technology in automation Application in printing products also continued to deepen. Based on machine vision technology, The online printing quality detection system in order to accurately and fastly extract the moving target(Yin shuapin) from the collected image without miscellaneous scene, it applied a high-tech moving target detection technology which melting automatic control, image processing, pattern recognition, artificial intelligence and other different academic fields. To quickly and accurately extract the moving object from the complex background is a basis for further analysis and processing of printing image, it not only greatly improve the efficiency of detection and reduce the complexity of the algorithm. After further study of related theory of on-line defect detection system and video image moving target detection technology, we designed the solution of this system, approach the solutions as the following aspects:(1) The original image preprocessing is an essential part in the segmentation of moving objects in the digital image processing technology, this technology mainly includes the denoising, binaryzation, morphological processing, grayscale, histogram processing and so on. Smoothing technique can remove the noise and unstable factors in the image, according to the type of noise, our use the median filter to remove denoising; The selection of threshold is the key of binaryzation, it directly affects the image segmentation, we contrast and analysis two kinds of global and local threshold methods, finally adopts the global threshold algorithm of Otsu threshold selection; because of environmental factors the target inevitably exist empty or fracture, so it requires the use of morphological opening and closing operation of the access to the moving target processing, makes the target more clear and soft. So it requires the use of opening and closing operation in morphological processing for moving target to make the target more clear and soft.(2) Considering the rate of unqualified prints is relatively small in printing process, In most cases the prints are qualified, and in other side, take real-time into consideration, so we decide to set of random operator randomly extracted image frame, and then extraction and detection.(3)Research and analysis of the moving target detection algorithm on current several better algorithms, we designed a three frame difference method conbine the Gauss algorithm to extract target. This method is not only solve the three frame difference method’s disadvantage which cannot extract the target inaccurate but also overcome the background subtraction of light sensitive weakness, and using the MATLAB R2008 a to verify the effectiveness of the algorithm.