Image Registration And Defect Inspection Algorithms And Their Application In Printing System

At present, most of the small and medium domestic printing companies urgently need automatic inspection system for inspecting prints defects, which is based on machine vision and image processing theory. However, domestic research in this area has just begun. To improve the overall level of China's printing industry, it is important to carry out the related research. This thesis mainly studies the following contents:First, the innovation of this thesis is to propose a new image inspection method based on the human visual characteristics. This method first uses regional image segmentation to zone various defects. After that, use mathematical morphology to analyze the characteristics of defects, including defect size, type, area and other information. Finally, evaluate defects according to human visual characteristics, which can determine the defect impacting degree. Experiments show that the method has good artificial intelligence and be able to meet the demand of inspection system. Second, image pre-processing and defect inspection algorithms are the cores of the printing defect inspection system. This thesis first analyzes the several image pre-processing problems, including the conversion of the color printing image to gray image, noises filtering, image registration, obtaining difference image, and proposes corresponding solutions for the above problems. Image registration is very important and difficult. So according to the demands of the system and the in-depth comparative study for the commonly used image registration algorithms, we explore a suitable registration algorithm for inspection system, which is the combination of pyramid and template matching image registration algorithm. Experiments show that the method not only has high registration accuracy, but also has high registration rate, which can meet the requirements of the inspection system.Third, according to the inspection system functional requirements provided by printing company, we design specific system function modules, and realize the software part of inspection system in the VC++6.0 and OPENCV development environment. The software part can supports the input of printing image, image pre-processing, defects inspection, list output for defects, human-computer interaction and so on.