Xianyi Zeng

Nonwoven uniformity identification using wavelet texture analysis and LVQ neural network

In this paper, an approach to grade nonwoven uniformity by combining wavelet texture analysis and learning vector quantization (LVQ) neural network is proposed. Six hundred and twenty-five nonwoven images of five different grades, 125 images of each grade, are decomposed at four different levels with five wavelet bases of Daubechies family, and two kinds of energy values L^1 and L^2 extracted from the high frequency subbands are used as the input features of the LVQ neural network solely and jointly. For each grade, 60 comparative experiments are employed to evaluate the performance of our method, which takes into account three effect factors, wavelet base (the length of filter), decomposition level and feature set. Experimental results on the 625 nonwoven images indicate that just use L^1 as feature calculated with db6, at level 3, the identification accuracy of grade A, grade C and grade E are 100%. When the nonwoven images are decomposed at level 3, the minimal average identification accuracy of five grades with five different wavelet bases is 87.7%.

Selection of relevant variables for industrial process modeling by combining experimental data sensitivity and human knowledge

Selection of relevant variables from a high dimensional process operation setting space is a problem frequently encountered in industrial process modeling. This paper presents two global relevancy criteria, which permit to formalize and combine the sensitivity of experimental data and the conformity of human knowledge using a liner and a fuzzy model, respectively. The performances of these relevancy criteria and some well-known selection methods are compared through artificial and real datasets. The result validates the outperformance of fuzzy global relevancy criterion, especially when the number of learning data is small and noisy.

Wavelet energy signatures and robust Bayesian neural network for visual quality recognition of nonwovens

In this paper, the visual quality recognition of nonwovens is considered as a common problem of pattern recognition that will be solved by a joint approach by combining wavelet energy signatures, Bayesian neural network, and outlier detection. In this research, 625 nonwovens images of 5 different grades, 125 each grade, are decomposed at 4 levels with wavelet base sym6, then two energy signatures, norm-1 L^1 and norm-2 L^2 are calculated from wavelet coefficients of each high frequency subband to train and test Bayesian neural network. To detect the outlier of training set, scaled outlier probability of training set and outlier probability of each sample are introduced. The committees of networks and the evidence criterion are employed to select the most suitable model, given a set of candidate networks which has different numbers of hidden neurons. However, in our research with the finite industrial data, we take both the evidence criterion and the actual performance into account to determine the structure of Bayesian neural network. When the nonwoven images are decomposed at level 4, with 500 samples to training the Bayesian neural network that has 3 hidden neurons, the average recognition accuracy of test set is 99.2%. Experimental results on the 625 nonwoven images indicate that the wavelet energy signatures are expressive and powerful in characterizing texture of nonwoven images and the robust Bayesian neural network has excellent recognition performance.

Construction of a prediction model for body dimensions used in garment pattern making based on anthropometric data learning

Abstract Using artificial intelligence to predict body dimensions rather than measuring them physically is a new research direction in apparel industry. If implemented, this technology can reduce costs and improve efficiency. In this paper, we proposed a back propagation artificial neural network (BP-ANN) model to predict pattern making-related body dimensions by inputting few key human body dimensions. In order to construct the proposed model, anthropometric measurements of 120 young females from the northeastern region of China were collected. The data were then used for training and the proposed model. The results showed that the prediction of the developed BP-ANN model is more accurate and stable than that of linear regression (LR) model. As great as the LR model was at pattern making, the BP-ANN model is even better. In the future, the precision of the proposed model can be further improved if the size of the learning data increases. The proposed method can be especially useful in making garment pattern for form-fitting clothing.

The visual quality recognition of nonwovens using a novel wavelet based contourlet transform

In this paper, a novel wavelet based contourlet transform for texture extraction is presented. The visual quality recognition of nonwovens based on image processing approach can be considered as a special case of the application of computer vision and pattern recognition on industrial inspection. For concreteness, the method proposed in this paper can be divided into two stages, i.e., the feature extraction is solved by wavelet based contourlet transform, which is followed by the grade recognition with support vector machine (SVM). For the texture analysis, we propose a novel wavelet based contourlet transform, which can be considered as a simplified but more sufficient for texture analysis for nonwoven image compared with version of the one introduced by Eslami in theory view. In experiment, nonwoven images of five different visual quality grades are decomposed using wavelet based contourlet transform with ‘PKVA’ filter as the default filter of Laplacian Pyramid (LP) and Directional Filter Bank (DFB) at 3 levels firstly. Then, two energy-based features, norm-1u2009L1 and norm-2u2009L2, are calculated from the wavelet coefficients at the first level and contourlet coefficients of each high frequency subband. Finally, the SVM is designed to be a classifier to be trained and tested with the samples selected from the feature set. Experimental results indicate that when the nonwoven images are decomposed at 3 levels and 16u2009L2s are extracted, with 500 samples to train the SVM, the average recognition accuracy of test set is 99.2xa0%, which is superior to the comparative method using wavelet texture analysis.

Identification of nonwoven uniformity using generalized Gaussian density and fuzzy neural network

A joint method to identify nonwoven uniformity by combining wavelet transform, generalized Gaussian density (GGD) and generalized dynamic fuzzy (GDF) neural network is presented in this paper. Six hundred and twenty‐five nonwoven images of five different grades, 125 images of each grade, are decomposed at three different levels with coif4 wavelet base. Wavelet coefficients in each subband are independently modeled by GGD model, while the scale and shape parameters of that are extracted as input features of GDF neural network. For comparison, two energy‐based features are also extracted from wavelet coefficients directly, the number of which is the same as the scale and shape parameters estimated from GGD model with maximum likelihood (ML) estimator. Experimental results on the 625 nonwoven images indicate the GGD model parameters are more expressive and powerful in characterizing textures than the energy‐based ones. The proposed method has high identification accuracy, such as when the images are decomposed at Level 3 and described with GGD model parameters, the identification accuracies of five grades are all 100%. Additionally, to reduce the redundancy of the generated fuzzy rules, an effective complementary approach, fuzzy rule base reduction based on ‘CityBlock’ distance is proposed.

Visual quality recognition of nonwovens based on wavelet transform and LVQ neural network

An approach to identify visual quality of nonwoven products by combining wavelet transform and learning vector quantization (LVQ) neural network is proposed in this paper. 625 nonwoven images of 5 different visual quality grades, each including 125 images, are decomposed at four different levels using five wavelet bases of the Daubechies family. The energy values L2 extracted from the high frequency subbands are used as the input features of the LVQ neural network. In our research, comparative experiments are employed to evaluate the performance of the proposed method, which takes into account three effect factors, including the wavelet base (the length of filter), the decomposition level and the size of training set. Experimental results show that this approach can lead to high degree of success rate in nonwoven visual quality recognition.

A Human Perception-Based Fashion Design Support System for Mass Customization

When developing mass customized products, human perception on products, including consumer’s and design expert’s perception, should be integrated into the process of design. So in this paper, we originally propose a fashion decision support system for supporting designer’s work. In this system, we first characterize and acquire fashion expert perception and consumer perception on human body shapes. Next, these perceptual data are formalized and analyzed using the intelligent techniques, such as fuzzy set theory, decision tree, and fuzzy cognitive map. The complex relations between these perceptions as well as the physical measurements of body shapes are modeled, leading to the criteria which will permit to determine if a new design style is feasible or not for a given fashion theme. The proposed system is aimed to support customized design and mass market selection in practice.

Visual quality recognition of nonwovens using generalized Gaussian density model and robust Bayesian neural network

This work is dedicated to develop an algorithm for the visual quality recognition of nonwoven materials, in which image analysis and neural network are involved in feature extraction and pattern recognition stage, respectively. During the feature extraction stage, each image is decomposed into four levels using the 9-7 bi-orthogonal wavelet base. Then the wavelet coefficients in each subband are independently modeled by the generalized Gaussian density (GGD) model to calculate the scale and shape parameters with maximum likelihood (ML) estimator as texture features. While for the recognition stage, the robust Bayesian neural network is employed to classify the 625 nonwoven samples into five visual quality grades, i.e., 125 samples for each grade. Finally, we carry out the outlier detection of the training set using the outlier probability and select the most suitable model structure and parameters from 40 Bayesian neural networks using the Occams razor. When 18 relevant textural features are extracted for each sample based on the GGD model, the average recognition accuracy of the test set arranges from 88% to 98.4% according to the different number of the hidden neurons in the Bayesian neural network.

Garment Fit Evaluation Using Machine Learning Technology

Presently, garment fit evaluation mainly focuses on real try-on and rarely deals with virtual try-on. With the rapid development of e-commerce, there is a profound growth of garment purchases through the Internet. In this context, fit evaluation of virtual garment try-on is vital in the clothing industry. In this chapter, we propose a Naive Bayes-based model to evaluate garment fit. The inputs of the proposed model are digital clothing pressures of different body parts, generated from a 3D garment CAD software, while the output is the predicted result of garment fit (fit or unfit). To construct and train the proposed model, data on digital clothing pressures and garment real fit was collected for input and output learning data, respectively. By learning from these data, our proposed model can predict garment fit rapidly and automatically without any real try-on; therefore, it can be applied to remote garment fit evaluation in the context of e-shopping. Finally, the effectiveness of our proposed method was validated using a set of test samples. Test results showed that digital clothing pressure is a better index than ease allowance to evaluate garment fit, and machine learning-based garment fit evaluation methods have higher prediction accuracies.