Research On Key Technologies Of Printing Roller Defect Detection Based On Visual Saliency

Print quality has always been the focus of concern in the printing industry,and its quality inspection has been throughout the printing process,because the printing roller is the key equipment in the entire print production process,and the printing roller surface reflectivity is high as well as the printing roller surface by the interference of the processing environment,will lead to the production of printed products inevitably certain problems,which affects the quality level of printed products,bringing a great deal of production companies economic loss and loss of reputation.With the gradual development of the target detection field,how to accurately and quickly extract the target and solve the problem of false edge weak edge gradually becomes a research topic of visual saliency target detection.To address this problem,the research in this thesis is as follows.(1)In response to the demand for high accuracy and high efficiency of printing roller surface defects varying and roller defect detection,a printing roller surface defect detection algorithm combining improved frequency-tuned saliency and threshold segmentation is proposed.Firstly,the image is pre-processed by non-local mean filter to solve the influence of noise on the image;then the Lab color space of frequency-tuned saliency algorithm is converted to HSV color space,and each feature value is normalized and weighted separately to extract the primary saliency map;finally,the saliency map is binarized by using iterative threshold segmentation algorithm to obtain the final defect image.(2)For the complex surface state of the printing rollers and the problem of large interference of the background texture of the rollers,this study introduces a saliency detection algorithm based on global contrast saliency.Firstly,removing uneven light from the printing rollers by leveling the light;secondly,the global contrast saliency detection algorithm is used to weaken the surface texture of the rollers and obtain the primary saliency map;finally,the defect boundary extraction in the image is combined with Canny operator and multi-scale detail enhancement algorithm.(3)Based on the actual industrial environment conditions and the prominence defect detection algorithm of the previous research,a printing roller surface defect detection system including hardware module and software module and integration is built.The functions of roller defect acquisition,roller motion control and roller defect detection are realized to meet the actual engineering application conditions.