Automatic Color Separation And Repeat Pattern Detection Of Printed Fabric

Colour and design are two important characteristics of printed fabric appearance, which not only embodies the value of the printed fabric, and comparing the two important features of different fabric. Because color printed fabric patterns are a repeating pattern of the use of printing equipment to print a color pattern and then be cased form, so repeating pattern separation and detection is crucial two analytical projects in the printing textile printing processes. Currently, many image processing and analysis techniques are widely used in the analysis of the fabric, but most are used to analyze the gray image, and the study of color fabric image has in the development stage. With the rapid development of computer technology, computer-assisted separation systems have been applied to fabric printing process color separations, although compared to the traditional manual operations has improved over time, but the system is not only expensive and still artificial follow-up operation, the only complete secondary separation cannot really realize full automation and can not detect a repeating pattern of the fabric. Therefore, the study of printed fabrics automatic analysis and detection systems in the current textile printing and dyeing industry, the necessity and urgency is obvious.Automatic color separation of printed fabrics and repeating pattern detection algorithm is the core of the automated detection system in this article. The main contents and work as follows:(1) SOM neural network combined with K-means clustering are modified algorithm is used to implement color fabric automatic color separations: the system by using genetic algorithm to search the optimal solution of characteristic to get the same as the original image color distribution of species and amount of sub image is used for colour separation.On the color separation to achieve the two layers of clustering method: the first layer, using SOM neural network to outnumber the clustering number of initial clustering samples, the samples with the same or similarity qualitative characteristic vector into the same category.Second, because in the feature space, according to the distance between the neurons or connection weights of features, such as by K-means algorithm are improved after training on the SOM network competition layer of neurons to conduct a clustering, thereby completing the color separation.(2) For pattern regularity of repeat fabric, improved template matching method is used for the test.The algorithm is conducted on the basis of traditional template matching method is improved, with appropriate size of the area instead of the traditional algorithm of each pixel match, with the characteristics of the regional code instead of each pixel similarity determination.The regional characteristics of coding, when certain areas in the image to be detected with the characteristics of the template image matching block number is greater than a certain threshold, is that the image to be detected in the region is matched with the template.The algorithm not only has high matching accuracy, but also the matching time significantly reduced, solve the problem of the traditional matching algorithm long-running time.(3) For fabric repeating patterns change size, brightness or point of view, an algorithm based on SIFT feature matching is used in the detection.This algorithm by gaussian pyramid and the establishment of the Do G scale space, accurate positioning feature point location and scale, determine the direction of feature points and generate 128-dimensional feature point descriptor and other process in the image size, brightness and Angle change, etc has good robustness and stability.Therefore, the algorithm is used to detect repeating patterns printed fabric, improved the disadvantages of template matching method, which cannot identify changes in the angle or size of the repeat printing design. For images with such changes it can have a good detection effect.