Study And Application On Web Inspection Method Based On Machine Vision

With the development of persistent domestic economy, the paper industry in china has changed fundamentally, and the production of paper and boxboard is increasing significantly. However, it is not easy to inspect the paper defects by manual inspection, which caused by equipments abrasion, quality of raw material and environment pollution. Therefore, this paper studies and discuses the web inspection method based on machine vision. And the system integration with the results of research is completed by existing equipments in the laboratory, by which, web inspection in papermaking can be simulated.According to the basic principles of machine vision, this paper studies the image processing algorithms for four types of common paper defects, including image smoothing, image enhancement, image division, and image mathematical morphology processing.Besides, ten image features are also extracted and analyzed in this paper, and the classifiers is designed by BP neural network, which has achieved good results of 91% on identification rate. And the hardware system is built by using intelligent image sensor, computer and other equipments. Finally, the web inspection system with image processing, feature extraction and classification is developed independently by Delphi, MATLAB and DVTSID control, which achieves the expected results.The method applied in the system still has its limits which has a long distance to the utility because of the manifold paper defects, although this paper has acquired some achievements. Therefore, this paper concludes deficiencies in research and gives suggestions for focus of next work.