Jayanta K. Chandra

Neural network trained morphological processing for the detection of defects in woven fabric

Basic morphological operations such as the erosion, dilation, opening, and closing often fail to detect various types of defects that may be present in woven fabric, mainly because of the heuristic selection of structuring element needed for these operations. In this paper, an artificial neural network (ANN) is utilized for the selection of structuring element, where ANN is trained by two pre‐assigned normalized numbers related to the warp and weft counts of the test fabric. The test gray fabric image is pre‐processed to remove noise and the interlaced grating structure of weft and warp and then converted to a binary image by thresholding. An intensity threshold value of the processed fabric image and the dimension of a sliding window needed for correlation operation are obtained from the trained ANN. Defects are detected after morphological reconstruction of the processed binary fabric image, where an ANN trained structuring element is used. The technique is tested on 317 samples for eight different types of defects in three types of plain woven fabrics from TILDA database and 92.8% success of detection is achieved.

Detection of defects in fabrics using subimage-based singular value decomposition

Matrix singular value decomposition technique is employed for the detection of defects in fabrics. Firstly, a region of interest (ROI) containing the defect is identified by a proposed adaptive partitioning technique – thus reducing the computational duty of operating over the whole image. The ROI portion of fabric image is then divided into small nonoverlapping subimages to further reduce the computational complexity and the average singular values of the subimages are calculated. To remove the interlaced warp–weft grating structure from ROI, which is the global information in the fabric image, selected singular values associated with positive average singular numbers are rejected and the fabric image is reconstructed to yield the image of the defect. Since the resulting image is saturated with noise and some unconnected parts mainly due to dissimilarity of the subimage of the fabric structure, postprocessing is carried out by binarization and edge detection to yield the edge map of the defect. Validity and feasibility of the proposed approach is established for detection of defects form images of TILDA database. The detection rate of 95% and detection success rate of 94.1% are achieved when tested over 460 samples.

Morphological Reconstruction Operation for the Detection of Defects in Woven Fabric

Basic morphological processing i.e. erosion and dilation, often fails to detect various types of defects that may be present in woven fabric, mainly because of the presence of interlaced grating structure of the weft and warp of the fabric. In this paper basic grating structure is optically filtered and then gray scale morphological reconstruction operations are applied for the detection of defects. The results have shown the usefulness of the proposed method during detections of various types of defects in woven fabric, as illustrated by the portions given in this document.

Class dependent 2d correlation filter for illumination tolerant face recognition

This paper proposes a class dependent 2D correlation filtering technique in frequency domain for illumination tolerant face recognition. The technique is based on the frequency domain correlation between phase spectrum of reconstructed image and the phase spectrum of optimum correlation filter. The optimization is achieved by minimizing the energy at the correlation plane due to resonstructed image and maximizing the corelation peak. The synthesis of optimum filter is developed by using the projecting image. Peak to side lobe ratio (PSR) is taken as the metric for recogntion and classification. The performance evaluation of this technique is validated by comparing performance of other unconstrained filtering techniques using benchmark databases (Yale B and PIE) and better results are obtained.

Detection of Defects in Fabric by Morphological Image Processing

Defects are generated in woven fabric due to improper treatments in weaving machines, spinning errors and inadequate preparations of fiber at the spinning stage. The economic viability of a weaving plant is significantly influenced by the extent of its success in eliminating defects in fabric. Detection of defects is generally carried out by time consuming and tedious human inspection. Such manual inspection procedures are commonly agreed upon to be inefficient with detection efficiency suffering from deterioration due to boredom and lack of vigilance. The problem is accentuated by the presence of several types of defects those may occur in woven fabric at random. In textile industry, imaging and image processing techniques are investigated for off-line and on-line visual inspection of fabric for the detection of defects (Zhang & Bresse, 1995; Drobino & Mechnio, 2006). The basic philosophy of detection of defects by such techniques is guided by the analysis of the image of fabric for distinguishing properties, those can be used to discriminate between defective and first quality fabric. In most cases, measurements are made on the first quality fabric and are then compared with the measurements made on the test fabric. Severe deviations in the measured parameters are used to indicate the presence of defects. Defects are then categorized into several types. However, the recognition of a particular type of defect amongst various classified types always remains a problem even in the context of presently available advanced image processing technology. Moreover, massive irregularities in periodic structures of woven fabric (particularly for fabrics manufactured from natural fibers) introduce very high degree of noise, which make identification and classification of defects difficult. The problem is accentuated very much due to the hairiness of natural fibers. Elaborate image processing algorithms are usually adopted for detection and recognition of defects (Sakaguchi et al, 2001). Recent reviews are available on various techniques, those can be applied for such tasks (Xie, 2008). In this chapter we are interested to explore one of such techniques which can be termed as morphological image processing, for the detection of defects in woven fabric. The techniques of morphological image processing are widely used for image analysis and have been a valuable tool in many computer vision applications, especially in the area of automated inspection (Haralick et al, 1987). Many successful machine vision algorithms used in character recognition, chromosome analysis and finger print classification are based

Sub image based eigen fabrics method using multi-class SVM classifier for the detection and classification of defects in woven fabric

Human visual system can identify larger defects taking place on the woven fabric. But it is very difficult to classify and identify the small fabric defects by a human inspector. In the textile industries the defect detection by a human inspector affects the production tremendously. Thus this paper gives a solution of this problem by developing an automatic fabric defect detection system, based on the computer vision. The sub image based PCA method is applied for the extraction of the feature from the training and test fabric images and the multi-class SVM classifier is used for carrying out the classification task. The method is tested on the standard TILDA database of fabric defect and a success rate of 96.36% is achieved.

Singular value decomposition method for the detection of defects in woven fabric refined by morphological operation

In this paper a new approach for the detection of defects in woven fabric is presented where the singular value decomposition (SVD) method is used. SVD basically removes the interlaced grating structure of the waft and warp of the fabric leaving aside the defective part of the fabric. An intensity threshold value along with the module of definite size is considered for the binarization of the background free fabric image. Finally, for the removal of the noise from the binary fabric image the morphological opening operation with the suitable structuring element is performed. The technique is tested on 287 fabric samples consisting of five different types of defects in three types of woven fabrics from TILDA database. 94.08% success rate of detection of defects is achieved.

Feature extraction and classification of woven fabric using optimized Haralick parameters: A rough set based approach

Classification of fabric samples into classes is highly required for automatic fabric inspection systems, as many of the fabric defects are defined relative to the fabric classes. The texture of the fabric surface is the best way to represent a fabric class, corresponding to which the statistical measures are the Haralick parameters. As all of the Haralick parameters are not responsible for fabric classification and there are no universal Haralick parameters for classifying all types of fabric samples, so it is necessary to determine a subset of Haralick parameters that gives best classification result for the fabric classes under consideration. This subset of Haralick parameters is termed as optimized Haralick parameters of the fabric classes under consideration, which has been determined by using the rough set theory. The developed system has been tested on TILDA database and its superiority with respect to the non-optimized Haralick parameters is established in terms of classification result and separability index.

Quantum signal processing-based visual cryptography with unexpanded shares

Abstract. This paper proposes a visual cryptography scheme (VCS) based on quantum signal processing (QSP). VCS is an image encryption technique that is very simple in formulation and is secure. In (k,n)-VCS, a secret binary image is encoded into n share images and minimum k shares are needed to decrypt the secret image. The efforts to encrypt a grayscale image are few in number and the majority are related to grayscale to binary conversion. Thus, a generalized approach of encryption for all types of images, i.e., binary, gray, and color is needed. Here, a generic VCS is proposed based on QSP where all types of images can be encrypted without pixel expansion along with a smoothing technique to enhance the quality of the decrypted image. The proposed scheme is tested and compared for benchmark images, and the result shows the effectiveness of the scheme.

Feature based optimal trade-off parameter selection of frequency domain correlation filter for real time face authentication

A real time fully automatic face recognition method is reported in this paper, where a frequency domain correlation filter known as optimal trade off maximum average correlation height filter (OTMACH) is optimized for real time application. The modification is done in selecting the optimal parameters for filter based on feature selection of face images by principal component analysis. The features are then used to train generalized regression neural network. The test face image is captured by a webcam camera interfaced with a PC and the recognition result is demonstrated by go-no go signals. The OTMACH filter as modified, gives better recognition rate compared to others conventional correlation filters. The system accepts reasonable variations in expression, pose and illumination conditions of face images.