Research Of PS Plate Surface Defect Detection Based On Image Processing

In the modern printing industry, Presensitized Plate (PS plate) plays an increasingly important role as the most famous offset printing plate, whose quality determines to a great extent the quality of the printing products. Since the surface quality of PS plate is the key factor of the whole performance of PS plate, the defect detection of PS plate is of great value to improve the quality of PS plate.The traditional artificially visual and electronic detection of PS plate can hardly meet the demands of modern detection manner with large-scale, high speed and automatic trend. So it is urgently needed to realize the automatic detection of PS plate. Nowadays, with the rapidly development of image acquisition equipments, computer software and the digital image processing technology, it is feasible to realize automatic surface defect detection of PS plate based on the technology of image processing and pattern recognition.In this paper, an automatic surface defect detection scheme of PS plate is proposed to meet the requirements of real-time, rapid and high-precision detection. First, analyze the image features of PS plate from the following three aspects: shape, edge and gray scale. After summarized and characterized the defect features of PS plate image, a suitable image processing algorithm flow is proposed as follows. In the image processing flow, an improved median filtering is carried out to get rid of noises. Then grayscale stretch is made to enhance the contrast between target image and background image. An improved iteration threshold method for defect segmentation is proposed, which uses iterative threshold segmentation after the image of PS plate is smoothed and has clear double apices.After the segmentation of defect image of PS plate, we choose major axis, minor axis, circumference, area, roundness and unbiased variance of gray as the characteristic parameters. A fast scanning algorithm is put forward to extract features of defects, which can avoid repeated scan of image so as to improve scanning speed. At last, a method based on regulations is put forward to realize the classification and recognition of white point, dirty point and scratch mark based on the extracted defect features.Finally, we choose Visual C++6.0 as the development platform to realize the above detection algorithms. For 1024~*762 pixels gray image of PS plate, it costs 55ms- 120ms to finish the detection and the detection precision is 0.2mm in diameter (including point defects and linear defects). The average accuracy reaches 96.7%.The experimental results show that the detection methods can realize real-time detection of PS plate under detection speed of 10 m/s, which can meet the detection requirements of industry.