Decision Fusion for Visual Inspection of Textiles

Abstract

Automated Visual Inspection (AVI) is the process of detecting, analyzing and classifying abnormal structures in a surface using machine vision techniques. The increasing competition in the industrial sector imposes high requirements on controlling the quality of flat surface products such as textiles, paper, steel slabs, glass, plastic films, foils, parquet slabs, ceramics etc. Automation of the visual inspection process saves companies a lot of time and raises the quality of their products by avoiding the subjectivity, boredom and slowness of the traditional human-based inspection process. When a defective product reaches the consumer, the company's reputation will suffer. AVI makes 100% quality control and documentation possible. The proposed decision fusion architecture is shown in Figure 1.The paper is organized as follows: the next section gives an overview of the existing AVI approaches. Following sections describe the role of homogeneity for faster defect detection; the adaptive window sizing; the feature sets used Abstract Defect detection of textiles is a necessary requirement for quality control and customer satisfaction. This paper presents a system for decision fusion in order to enhance the accuracy of defect detection in textiles. A multi-classifier decision fusion technique based on majority voting is presented to solve the problems of sensitivity to parameter variation and to make use of the advantages of the individual feature sets for accurate texture characterization. Features based on Gray Level Co-occurrence Matrix (GLCM), Laws Energy (LE) Filter, Local Binary Patterns (LBP), HU Moment invariants, Moment of Inertia (MOI) and Standard Deviation of Gray levels are used to train a set of Learning Vector Quantization (LVQ) classifiers. Detection accuracies of classifiers trained on single-feature sets are compared with those of the majority voting among the individual classifiers. The results obtained from majority voting indicate that the decision fusion technique improves the accuracy and reliability of the detection process. Empirical results indicate the high accuracy of the presented approach. The correct defect detection rate of the proposed approach is 98.64% with an average false acceptance rate of 0.0012.

Monitoring tool wear using classifier fusion

“…The fusion procedure for multiple classifiers [56][57][58] being considered here is illustrated schematically in Figure 2, using the hidden Markov (HMM), Bayesian rule, Gaussian mixture (GMM), and K-means models [59,60].…”

“…Since individual classifiers perform differently, depending on the type of application, the use of multiple classifiers has been investigated in some fields for making monitoring systems more robust [16][17][18][19][20][21][22]. Conceptually, this is similar to sensor fusion which capitalizes on the advantages of individual sensors and reduces sensitivity to noise [8,[23][24][25][26][27].…”

A simple computational framework for defect detection system with orthogonal polynomials transcoded coefficients

“…Hence automatic defect identification, with the help of digital image proceesing scheme, is a natural alternative. In this direction, defect detection in application areas such as automated manufacturing [2,3], textile fabric [4][5][6][7][8][9][10][11], film industry [12][13][14], wood [15][16][17], construction industry [18], Printed Circuit Board (PCB) ( [19][20][21], wafer [22][23][24][25], solar cells [25][26][27][28], paper industry [29], leather [30], food processing [31,32], and rails [33] are reported in the literature. These image analysis techniques, designed for defect detection, are implemented either on non-textured surface like paper and glass materials or homogeneously textured surfaces like textile or on structural patterns like semiconductor wafer dies and Liquid Crystal Display (LCD).…”

Homoscedasticity for defect detection in homogeneous flat surface products

“…Different approaches give different results that contribute to greater diversity of results and combinations of them will lead to better results than any individual approach. Tolba et al 17 presented a multi-classifier decision fusion technique based on majority voting to solve the problems of sensitivity to parameter variation and to make use of the advantages of the individual feature sets for accurate texture characterization. Mahajan et al 18 have concluded that the contextual-analysis-based defect-detection approaches are more promising than filter-based decomposition approaches.…”