Based On The Semi-circle Skirt Virtual Fitting Fabric Classification Of Support Vector Machine Modeling

A prominent problem to fashion design and manufacture is selecting the suitable fabric to perform the clothing shape, and creating the perfect combination between fabric and clothing shape. In order to solve this problem, we study the shape of half-circle skirts, 3D virtual try-on platform and Support Vector Machines (SVMs) are applied to construct the model for fabric classification, the physical and mechanical properties of fabric are used as the input of the classification model and the categories of half-circle skirts shape are used as the output of the model. With the model we can predict the categories of half-circle skirts shape and get the skirt feature, determine whether to choose this fabric for the half-circle skirts production, to direct us of fabric selecting. In addition, if we want to make a certain category of half-circle skirts shape, we can select fabric by considering the corresponding physical and mechanical properties of fabric of the category. The model also provides a feasible method for other clothing styles of the fabric choosing.By analysis of the feather of half-circle skirts shape, we defines 15 parameters for objective evaluation of half-circle skirt shape: front length, front width, side width, front spread angle, side spread angle, bottom contour, wave number, average peck radius, average trough radius, average peck angle, average trough angle, standard deviation of peck radius, standard deviation of trough radius, standard deviation of peck angle, standard deviation of trough angle, and the corresponding formulas are presented.According to the actual use of fabrics for half-circle skirts, we selected 55 different fabrics as the experimental samples. The Fabric Testing Kit (FTK) which matching the 3D virtual try-on platform V-Stitcher were used for testing the physical and mechanical properties of fabrics. After that we import these data into the platform to simulate the skirt shape, and extract the parameters for objective evaluation of half-circle skirt shape from the simulating image. By using the platform, we can save a lot of actual production costs, simplify image extracting process, and effectively avoid production errors and other human factors.Factor analysis was carried out for the physical and mechanical properties of 55 fabrics, three main factors including thickness-weight and bending, weft stretching-bending and shear, warp stretching and bending were picked up, the accumulative contribution is 82.044%, their predictive equation were built up. Factor analysis was also carried out for parameters for shape objective evaluation of 55 half-circle skirts, four main factors including outline, peck line, tough line, detail were picked up, their accumulative contribution is 83.291%, their predictive equation were built up too. Cluster Analysis was used in the research, the object is parameters for shape objective evaluation of 55 half-circle skirts, 55 samples were clustered into 5 different categories, by discussing the shape and fabric feature of each category, we give selection recommendations of each category of half-circle skirts shape.On the basis of the study of the Support Vector Machines and its application principle and method applied to data classification, the model was trained and tested with the Libsvm toolbox, simulated in MATLAB. In this classification model, the input vectors are the physical and mechanical properties of fabric, the output vectors are categories of half-circle skirts shape. Finally, C-Support Vector Machine and radial base kernel function are selected to build the model. The optimal value of parameters C andγare 1 and 0.125 respectively.Finally, we chose three fabric samples, which do not belong to the 55 experimental fabric samples, to check the application and validation of the classification model. The verification mainly has three ways: one approach is to see whether the measurements of the physical and mechanical characteristic of fabrics belong to the fluctuation range of SVM model predicted. Another approach is to compare the index of the V-Stitcher simulate model with the index scope of the SVM model predict. The last approach is making a comparison between the feature of the actual half-circle skirts with that of the predicted one. The mainly function of this model is predicting the category of half-circle skirts based on the physical and mechanical characteristic of the fabric, and choosing suitable fabrics belong to the skirts shape categories of the model according to the designed half-circle skirts.