Analysis On Image Of Paper Defects Based On Machine Vision

With the rising of modern paper machine speed and width, the paper disease risk is also increasing. However, in the production process, it is not easy to inspect the paper defects by manual inspection. In order to improve the quality of paper sheets and production efficiency, the online paper defects detection system based on machine vision is required and imperative to replace the traditional manual inspection. Machine vision is based on image processing and analysis technology. This paper studied paper defect image processing and recognition algorithms according to the paper image features, and realized the detection of hole, spot, wrinkle and crack.This paper focuses on algorithms of paper defects image processing, features extraction and recognition of paper defects, and the software design of web inspection system. Image processing includes image smoothing, edge detection, mathematical morphology processing and threshold segmentation. By comparing the typical edge detection operator:Roberts operator, Sobel operator, Prewitt operator, Log operator and Kirsch operator and according to the disadvantage of the current operator, a new algorithm is brought out to detect paper defects edge. The research results show that the new algorithm is superior to traditional edge detection operator, in edge localization strengths, especially when detecting targets with small gray step. The mathematical morphology processing is used to further optimize the defect edge. In order to better extract the feature of every defect, this paper puts forward a new method of feature extraction based on paper defect areas. The grayscale, morphology and texture characteristics of four paper defects have been studied respectively, and the grayscale average, circularity, aspect ratio, rectangularity and relevant standard deviation are chosen as the features to distinguish the four paper defects. Finally, the BP neural network is used to design the paper defects classifier, and the experiment results indicate the approach gets a good performance.Finally, matlab2012a, Visual C++2010and SQL Server2008are used to write the software of web inspection system. The test results show that this software can detect the paper defects information for papers with more than one kind of defect, especially for hole, spot, wrinkle and crack.