| With the improvement of our living standard, people pay more attention tothe quality of the presswork. However, producers have to consider t he productionat the same time, so it’s a challenge for producers. This situation demands aefficient and accurate defect detection method for the presswork. The foreignproducers have made a lot of achievement in this respect. However, to us, theirequipment is expensive and has a high maintenance cost because of the technicalmonopoly. Since the year of2000, domestic defect detection method has made alot of progress, even has a lot of disadvantages compared with the foreignproduction.Focus on the above situation, in this thesis, we study the online detection ofprinting defects. Based on predecessor’s foundation, to raise the speed ofdetection, we mainly do the following study. We study the basic defects detectingmethods to design a new efficient algorithm in the condition of guarantee theaccuracy. What’s more, using the experiment platform, we conduct a lot ofexperiments based on PC and DSP, which helps to improve engineeringapplication.We study the basic detection methods in the printing defects detection. Themethods include two aspects, registration and detection. Aiming at imageregistration, we study three algorithms based on the image gray, transformdomain and image feature. In the respect of detection algorithms, we study someclassic algorithms based on pixels, layered detection and area projection.After the study, and taking the index of this thesis, we design a defectdetection method, the image registration uses two templates of designated areasand the detection is based on pros and cons of differential. The image registrationalgorithm is based on template matching, and the detection part develops fromdirect comparing each pixel of two images.Using the experimental platform with image gathering module, mechanicaldevice and motor control module, based on PC processor, and simulating theplanographic printing, we conduct a lot of experiments to analyze the performance of the method we design.To further improve the process of engineering application, we conduct thestudy of realizing the algorithm on the platform of DSP. And on the platformconsist of the DVS6446development board and Seed560USBplus emulator, weimplement our method basically. |