Research On Rapid Detection Algorithm For On-line Paper Defect

Papermaking industry has close relation with national economic construction and socialist construction. The quality of the paper directly determines the survival of papermaking industry. To detect out paper real-time defect accurately becomes the key technoLogy in papermaking industry. It is more and more impossible to inspect the paper defects by manual inspection because of the increasing web width and machine speed. Detection system with fast and accurate detection algorithm is the best choice to realize the professional paper defects detection.In order to obtain a real time detection system with fast and accurate detection algorithm, this paper builds a paper defect detection test platform. In terms of image processing, currently algorithm used on paper defect detection is introduced according to the defect paper image processing steps. Firstly, this paper introduces the image denoising algorithm which is the first step of image processing. Secondly, we introduce the image segmentation algorithm, which is an important image processing algorithm including regional segment and edge detection. We can distinguish target and background and classify the paper defects preliminarily only after segmentation. Later, in order to make up the error made by pure segment, this paper also introduces the application of mathematical morphoLogy in the inspection on paper defects. This article improves detection speed from detection algorithm and disposal of the web images. According to dark spots and holes'characteristics, this paper introduces double threshold segmentation method in which global threshold combines dynamic adaptive threshold. This method realizes parallel processing of detecting out the holes and dark spot with good effect. Besides, improvements to some algorithms for fast realization are introduced.This article dose a detailed study of wrinkle which is a complex paper defect and may occur in paper making. In view of the Hough transformation advantage on line testing,this paper uses Hough transformation to inspect wrinkle, gives the detection steps, and also introduces some fast Hough transformation method. Considering the poor effect of Multi-line Detection, this article introduces an original image erasure way used to Multi-line detection. Using this method on wrinkle detection achieves a big success.Based on the algorithms study, we use Visual C++ 6.0 for the software development platform. With industrial control computer for host computer, this paper designs an on-line experiment inspection system including image collecting, image classification, image characteristics marking, alarm, real-time image show, simulation image show and history data querying and so on. All the algorithms used are included in this system. Additionally, the system performance is tested and analyzed. Finally, for the future researchers, this paper makes some suggestions due to the deficiencies existed in this paper.