The Research Of Image Recognition Algorithm Used In Defect Inspection Of Large Format Screen Plates

As the development of silkscreen and the high price of screen plates, more and more companies turn their attention to the quality of screen plates, while the traditional method of human detection can not meet the demands .There isn't any automatic detection equipment for screen plates in our country, and some fancy equipments for printing are added detection function in other countries. But most of them only fit for PCB, which are expensive and small in size. That can not tally with the actual situation. The purpose of this paper is getting software solution for defect inspection of screen plates by doing some research in image recognition used in it.The paper described the structure of whole system, analyzed the given performance the system must reached, and then made hardware solution and software solution of the vision system. Image segmentation must be carried out by template matching because of the screen plates' feature. The original images can be fallen into two broad categories based on the status of mesh, either of them has its own solution. Plot and number the reference image before detection according to the moving of the collecting device. In this way, not comparing the image with the whole reference image but reading the relevant image larger than it while positioning, which can reduce search space and increase the speed of positioning.Otherwise, data processing must be finished as soon as possible, but the living algorithm of image matching can not meet the requirement. Considering the feature of original image and reference image ,the author analyzed advantage and disadvantage of existent algorithm, and then put forward a new algorithm based on projection with adaptive threshold. The experiment result showed that the improved algorithm is preferable on rapidity and accuracy.The defect is located by two steps. Most normal image can end its detection by the first step, and only the minority needed the second step. By this mean, miss ratio is reduced while the speed of detection is increased.Start the equipment of defect inspection of screen plates, and the result showed that the solution and algorithm put forward in this paper are better than human detection on rapidity, accuracy and stability.