The Study Of Detection Algorithm Of Paper Defects Based On Machine Vision

With the increased speed of paper machine, the possibility of producting the paper surface defects on the production line is also increased. At the same time, with the increasing abundance of material life, people's requirements about the quality to the paper is also improved accordingly. Furthermore, it has become the development trend to detect the paper defects using machine vision technology. Therefore, it is very necessary to detect the defects accurately for controlling the quality of paper in the process of production.Machine vision and digital image processing technology have been utilized, and some related key technology of paper defect detection has been studied, such as image filter and image edge detection algorithm, considerring current paper defect detection methods have many problems, for example, anti-noise performance is not good, detection accuracy is not enough and so on. CB morphology, the bit-plane decomposition technique and RPCA method have been combined, A paper defect detection algorithm based on CB morphology and gray code decomposition and a paper image segmentation algorithm based on RPCA are studied. The main contents of this paper can be summarized as follows:(1) The filter algorithm based on multi-scale CB morphological are studied. On the basis of learning common image filter algorithm, the filter algorithm based on multi-scale CB morphological is studied. Combined the advantages of multi-scale structure elements and CB morphology, this algorithm can weaken the dependence of the morphological results on structural elements. And the noise in image can be separated from image points effectively by the principle of "extension". Finally, the related filter algorithm are evaluated from two aspects of the qualitative and quantitative. Analyse shows that, the filter algorithm used in this papper can get better filter effect.(2) The edge detection algorithm based on CB morphology of multi-structure elements is studied. Common edge detection operators and edge detection algorithm based on morphology have been studied, and these algorithms have been simulated and analyzed. By selecting multi structure elements and combined CB morphology, the edge detection algorithm based on CB morphology of multi-structure elements is studied. This method can effectively avoid the problem that the effect of edge detection algorithm based on morphology is not good when the texture or edge of an image becomes more complicated. Qualitative and quantitative analysis show that, the edge detected by the edge detection algorithm based on CB morphology of multi-structure elements is more complete and clearer, more accurate positioning, and has better anti-interference.(3) The paper defect detection algorithm based on CB morphology and gray code decomposition is studied. When researching the existing paper defect detection algorithms, it is found that the large amount of image information makes the calculation of the algorithm increased, and thus the real-time performance of the algorithm is also affected. Considering on this, a paper defect detection algorithm based on bit plane is studied. In the new method, the paper image is decomposed by the means of bit plane, and the computation of the algorithm can be effectively reduced by choosing a bit plane for defect detection. At the same time, the bit plane expressed by gray code can alleviate the interference of detection caused by the tiny change of the pixels. Combined this methods with the above mentioned filter and edge detection algorithm, the paper defect detection algorithm based on CB morphology and gray code decomposition is studied. Simulation and analysis shows that, the algorithm can detect the defect better and has a certain effectiveness.(4) The paper image segmentation algorithm based on RPCA is studied. According to the linear correlation or similar characteristics between the pixels of image background, it is found that the matrix of paper defect image can be decomposed into sparse matrix and low rank matrix by using RPCA method, and the low rank matrix corresponds to the content of the background image, while the sparse matrix corresponds to the image which contains defects. In the following detection, just selecting the image corresponded by the sparse matrix for detection can meet the requirements of detection, and the amount of computation can be reduced effectively. The simulation result shows that this algorithm can be used for the segmentation of paper image and has good segmentation performance.