Application Of Image Segmentation And Template Matching In Cloth-cutting

There is a huge demand in textile industry that is to cut the various clothing patterns. The traditional approach which using the hand-cutting can hardly be satisfactory because of the low efficiency and the difficulty in normalizing the cutting-rule. To solve this problem, a new scheme is proposed in this paper: get the cutting-data by using image processing technique on cloth images, and then use the laser cutting machine to do the cutting job.The paper is divided into three parts:In the first part, an advanced region growing algorithm is proposed to make segmentation of the cloth images. It used scan line in the process of growing, by which much time is saved while the result remains as well as common region growing algorithm.In the second part, we designed the method to extract cutting-data from the "contour-pattern" images after the segmentation. In this sector, boundary tracking and boundary splicing is used to get the pixels sequence which is divided into several sub-paragraph by calculating the weight called "TurnValue" for all pixels. Then we make the fitting of the sub-paragraph to get the final cutting-data.In the third part, template matching is used to extract cutting-data from the "template-pattern" images after the segmentation. A new algorithm called "Multi-Direction Template Matching" is presented here, and is used to locate the template image in the scene image. The new algorithm is rotation-invariant and zoom-invariant.Experimental results are presented in all chapters to demonstrate the efficiency of the methods proposed in this paper. Summarization and the author's idea to extent study are given in the last chapter.