Research On Pattern Reuse Technology Based On Style Structure Feature Matching

Apparel pattern information as a technical document to guide the production of enterprises,has a vital role.The pattern maker must be familiar with the process of playing CAD software,pattern making,grading,marking,fitting and adjusting garment,master the entire the production scheduling of the garment factory.The quality and comfort of the pattern depend largely on the experience of the pattern maker.The production type of garments has shifted to small batches,multiple varieties,and short cycle,thus placing higher demands on the rapid response capability of enterprises.However,at present,there is a lot of repetitive work in the clothing industry.The reuse of the pattern mainly depends on the memory of the technicians from a large number of pattern libraries,and there is no clear focalization.Pattern reuse technology based on style structure features can be used to retrieve past data in a targeted manner,thereby obtaining pattern information of similar styles,reducing the requirements for clothing pattern maker,and improving the rapid response capability of enterprises.In this paper,the dress category is taken as the research object.According to the characteristics of the style structure and the pattern maker research,the outer contour of the dress is taken as the main research aspect,and the retrieval method based on the structural feature recognition and matching of the dress style is proposed.The main research contents and methods are as follows:(1)Determination of dress samples and contour extractionAccording to the analysis of the structural features of the style,there are 10 key points in the shoulder,waist,hip,mid-hem and hem.The program written by MATLAB finally extracted the outline contour shape and coordinates of the key points of 459 dress styles from international clothing brands,and sorted the original clothing styles figures,outline contour shape and twodimensional contour points data files of all samples according to the silhouette.The sample data and dress contour shapes required for the research of this paper were obtained.(2)Contours classification based on neural network pattern recognitionThree description schemes of dress contours have been developed,including width,height,values of angle cosine,and silhouette scale feature values.Combining the contours classification results with pattern recognition training,a better neural network model is obtained after the misjudgment samples are analyzed and adjusted.Among them,the first scheme has the best effect and the recognition accuracy is 93.5%,which can be used as a method for interactive silhouette classification.(3)Comparison of contour similarity ranking results of different algorithmsA total of 114 representative samples were selected from 459 samples.The contours similarity evaluation questionnaire was designed for the clothing engineering background personnel,and the similarity evaluation table of the dress style silhouette was obtained.The three algorithms of traditional Euclidean distance,Euclidean distance based on coefficient assignment and shape context matching are used to calculate the similarity of the dress contour shape.The accuracy and precision of the three methods are obtained by comparison with the similarity evaluation table.The Euclidean distance based on the coefficient assignment is guided by the similarity evaluation table,and the least squares solution is calculated to obtain the optimal coefficient.The matching result shows that the shape context algorithm has the best effect with an accuracy of 82.91% and an precision of 77%.(4)Main structural design of the pattern reuse systemThe application scenarios and requirements of the pattern reuse technology are mainly analyzed.The style structure feature elements are obtained through interviews with the pattern experts.The function of dress retrieval is preliminarily designed,and the overall framework and function modules of the pattern reuse system are further designed and divided.