The aim of this study was to propose a defect detection method of color printing image based on CIE L*u*v color space and SURF algorithm to solve rapid and accuracy problem under different light intensity. First, pre-treatment was conducted with median filter to eliminate noise of the color printing image. The RGB color space of both standard image and color printing image to be checked was translated into the CIE L*u*v color space. 122-dimensional features vectors of CIE L*u*v color space based on SURF algorithm were constructed. The parallel method of two threads was proposed to match among interesting points from standard image and color printing image to be checked with Euclidean distance. If interesting points of color printing image were all matched with interesting points of standard image, then the color printing image to be checked had no quality defect. Within range of 150~650 lux, this method can detect defects such as position offset, ink contamination, color change, etc., only taking 650 ms. This method of good adaptability to different light intensity can rapidly and effectively detect the defects of color printing image with single hardware configuration.
WANG Wen-ju, ZHAO Ping, CHEN Wei, XIE Han, SUN Liu-jie, JIANG Zhong-min, CHEN Jing-hua.
A Fast and Accurate Method of Defect Detection of Colour Printing Image[J]. Packaging Engineering. 2015(17): 112-118130