Research Of Paper Surface Defects Detection Algorithm Based On Machine Vision

At present, the majority of paper-making enterprises at home and abroad on thepaper defects generated in the production still use artificial way to detect and classify,nevertheless, which has greatly affected the production tasks, and also be formed easilythe problem of low detection accuracy and high failure rate. Therefore, research ofpaper surface defects detection algorithm based on machine vision not only canquickly and accurately achieve the paper surface defects detection and analysis,microscopic, visualization and automation, also for paper-making enterprises to reduceproduction costs and improve the detection efficiency, in short, it’s very meaningful toimprove the paper-making enterprise benefit and improve the paper-making processautomation.In this paper, the research background and development process of the webinspection system and paper surface defects detection technology is undertook in-depthstudy and master, and then paper surface defects detection algorithm based on machinevision is dissertated in detail. In view of the web inspection system in imageacquisition process encountered paper image quality problems, this thesis proposes amore advanced light adaptive compensation method to adjust the light sourceaccording to specific changes in the environment adaptive completed, which couldreal-time acquisition to the high quality of the image.On the basis of the successful acquisition of high quality paper images, based onlinear interpolation of the dynamic threshold selection algorithm is proposed in orderto achieve rapid detection of the holes, bright spots, dark spots, and dark spots underdifferent lighting conditions. Finally, the experiment indicated that this method todetect high-contrast paper defects is practical and effective, and can meet therequirements of the paper defect detection real-time and accuracy.At the same time, the paper also conducted in-depth research on complicatedpaper surface defects detection algorithm, in which the complicated paper surfacedefects such as wrinkle and crease has linear features, then there are a lot of on linedetection algorithm, including least squares linear fitting, Radon transform and Houghtransform etc. But the complicated paper surface defects which have close slope orintercept can not be distinguished effectively by above mentioned algorithm. Based onexisting algorithm, the Hough space was transformed into two-dimensional image, andthen extracted paper defects features, through which the wrinkle and crease surface defects could be detected more accurately. Finally, the experiment indicated that theimproved algorithm had higher detection precision and shorter time consumptioncompared with Hough transform.In order to further improve systemic of the paper defects detection, based onsupport vector machine in the paper defect detection and classification is studied.However, the research on multicategory classification support vector machine methodis still in its preliminary stage, although some progress has been made, there are manyaspects to be improved and researched.