| The color printing register marks detection system based on computer vision is used in on-line of off-line detection of the color register marks in the printing industry in the purpose of positioning the printing plate on-line or analyzing the printing quality off-line. In this text, algorithms such as preprocessing of color image, intensity segmentation, color segmentation, circle detection and abstraction of geometric characters are studied and a practical detection system consisting of both hardware and software has been set up.In the Preface, research background together with project target is presented. Also related content of computer vision is introduced. Three complementary colors and the color of black, i.e. the CMYK color model, are widely used in the color printing industry. The four colors are mixed up to obtain an arbitrary color. Because the mixing is done by printing the four colors at the same place in sequence instead of being done before the printing, the problem of registering occurred. If the four colors making up the desired color are not printed at the same place and at a distance that exceeds a predetermined value, the quality of the printing will not be guaranteed. In the printing industry, all kinds of register marks are used to decide if and with what accuracy the color printing is registered. At the present time, both on-line control and off-line analysis are performed manually and inefficiently. Accordingly, the task of register marks detection is proposed and studies on computer-vision-based system of detecting register marks are presented.In Chapter Two, algorithms based on image segmentation are studied. Those algorithms can be categorized into those based on intensity segmentation and those based on color segmentation. Segmented results will be the input to the post-processing. Pixels taken from the cross lines of the register marks are approximated into lines and the intersection points are calculated. Consequently the centers of the register marks are obtained. Algorithm of segmentation has a high demand for the quality of the original image and sensitive to the noise. Filtering is expected before segmentation.In Chapter Three, algorithm based on mathematical morphology are studied. This algorithm functions with an operator---a kind of operator on set theory, searching for the matched version in the original with the structural element. The key is to initialize the structural element. The parameters of the structural element is modified in accordance with the defined matching task. The... |
