Color Segmentation And Cyclic Primitive Segmentation Of Printed Fabrics Based On Clustering

The segmentation of fabric color and cyclic primitives has important value in the process of textile production and quality evaluation.The colorful printed fabrics are first formed by the designer who first circulates the segmented circular primitives on the computer into the pattern that the customer required according to a certain rule,and then prints it into a monochrome pattern by the machine,and then dyes them according to the results of the color segmentation.In addition,the evaluation of the color fastness of the fabric also requires color segmentation at first.In the traditional production mode,a large amount of manual participation is required to complete the division of fabric colors and cyclic primitives.This requires a lot of labor and is greatly affected by personal subjective judgments during the segmentation process,and the quality of segmentation is unstable.The existing color segmentation and cyclic primitive segmentation algorithms have not suitable for printed fabrics,insufficient color segmentation accuracy,difficult to accurately identify small-area colors,and the results of color segmentation do not meet the actual printed fabric production needs and do not link color segmentation and cyclic primitive segmentation.Therefore,this topic has carried out the following research with fabric color segmentation and primitive segmentation as the research content:(1)Color segmentation of printed fabrics: A color segmentation algorithm combining Self-Organizing Maps Neural Network(SOM)and Efficient Dense Subspace Clustering(EDSC)is proposed.The algorithm first uses the SOM algorithm to perform initial clustering on the pre-processed image data to reduce the amount of data of secondary clustering,and then uses the output of the SOM algorithm as the input of the EDSC algorithm for further clustering to complete the color segmentation.At the same time,the optimal SC is introduced in the clustering process of the EDSC algorithm to automatically determine the number of clustering centers.Finally,morphological operations are used to post-process the segmented sub-color pictures to make the color edges more suitable for actual production requirements.Experimental results show that the algorithm can accurately segment colors in complex printed fabric pictures,and the color edges of sub-color pictures are more in line with actual production requirements,and it can more accurately segment the small area color blocks in the fabric picture.(2)Cyclic primitive segmentation: Based on the sub-color image obtained by color segmentation,first perform edge detection to extract the pattern edges,and then perform hole filling and opening operations on the edge image to eliminate the adverse effects of incomplete patterns on the cyclic primitive segmentation,and then get the centroid coordinates by calculate,and the adjacent centroids' s difference between the horizontal and vertical coordinates is used as the primitive size of the sub-color picture,and then the cyclic primitive size of the fabric picture is determined based on the principle of containing the most information.Finally,we choose a location that closed to the center of the fabric picture for cyclic primitive segmentation.Experiments show that this method can accurately segment the cyclic primitives of the printing pattern.