The Research Of High Speed Inspection Method Of Printed Matter Based On Machine Vision

With the development of industrial automation, the speed of industrial products production becominghigher and higher.In the same time, consumer’s demand for product quality and consistency more and more strict. Especially in the printing industry,usually the printing speed can be upon to dozens or even hundreds of meters per minute, whilethe printing accuracy needs to reach submillimeter level. Which put forward high demands for thedefect inspection of printing products. On one handthedetecting speed should be very high that reach the speed of printing, on the other handthe detecting precision and accuracy should meet the quality requirements. To achieve the goal, the designer should use efficient algorithms of digital image processing, in the meanwhile a fine structure ought to be designed to make full use of the computer’scomputing power. This paper is based on the color bond pipeline renewal project of Tianjin re lian coated steel co., LTD. In addition to the study of detection algorithm, this paper also introduced many system development issues.Such as the control of the camera, the system architecture design, the system interface design, the storage and queryresults, the maintenance and upgrade of the system, etc.In this paper, the print fast on-line defect detection method is discussed from several aspects, both from the view point of system design, such as the selection and installation of hardware, the design patterns of system software, etc. And from the view point of image processing algorithms, such asthe algorithm of template matching, thealgorithm of defect detection, etc.From the perspective of system design, for the hardware system, this paper accomplished the selection and installation of the components. Such as choose the camera’s minimum horizontal frequency, minimum resolution, determine the installation position,etc.according to the actual demand. For the software system, according to the requirements of project,designed the architecture of the system to make the system easier to maintain.Andusing Qt for graphic user interface design, using theCameraLink serial communication function of the acquisition card to control the camera, usingthe multithreading technology to realize the image acquisition process, image processing process and image storage process asynchronous execution, using the MySQL database to supervise and control the data.From the perspective of algorithm design, this paper studied multiple template matching methods, and mainly introduced the methodof template matching based on random trees. This approach makes objects which are in the presence of perspective transformation, can still be accurate matched. And also designed a defect detection method based on template matching, whichcompares the qualified product model and product to be detected by different threshold in different areas so as to obtain the defect information.Finally, accomplished the system’s functional verification and show a method of software release.