Jaume Escofet

Detection of local defects in textile webs using Gabor filters

A method of image analysis is proposed for detection of local defects in materials with periodic regular texture. A general improvement and enlargement of vision system capabilities for versatility, full automa- tism, computational efficiency, and robustness in their application to the industrial inspection of periodic textured materials is pursued. In the pro- posed method, a multiscale and multiorientation Gabor filter scheme that imitates the early human vision process is applied to the sample under inspection. The designed algorithm automatically segments defects from the regular texture. A variety of examples of fabric inspection are pre- sented. In all of them defects are successfully segmented from the tex- ture background.

Modeling of woven fabric structures based on Fourier image analysis

The periodic woven structures of fabrics can be defined on the basis of the convolution theorem. Here an elementary unit with the minimum number of thread crossings and a nonrectangular two-dimensional comb function for the pattern of repetition is used to define woven structures. The expression derived is more compact than the conventional diagram for weaving, and the parameters that one needs to determine a given fabric can easily be extracted from its Fourier transform. Several results with real samples of the most common structures-plain, twill, and satin-are presented.

Fourier-domain-based angular correlation for quasiperiodic pattern recognition. Applications to web inspection

A Fourier-domain-based recognition technique is proposed for periodic and quasiperiodic pattern recognition. It is based on the angular correlation of the moduli of the sample and the reference Fourier spectra centered at the maximum central point. As in other correlation techniques, recognition is achieved when a high correlation peak is obtained, and this result occurs when the two spectra coincide. The angular correlation is a one-dimensional function of the rotation angle. The position of the correlation peak indicates the rotation angle between two similar patterns in the original images. Some optimizations for the discrete calculation of the Fourier-domain-based angular correlation are also proposed. Some applications of this technique to web inspection tasks, such as pattern recognition and classification, damaged web evaluation, and detection of defects, are presented and discussed.

Weave-repeat identification by structural analysis of fabric images

Two descriptions of the image of a web structure, a convolution model and an additive model, in both the spatial and frequency domains, are combined in the design of a method to extract information about the fabric structure by image analysis. The method allows the extraction of the conventional and also the minimal weave repeats, their size in terms of number of threads, their interlacing patterns, and their patterns of repetition. It is applicable to fabrics with square and nonsquare conventional weave repeat. Experimental results with image of real samples are presented and discussed.

Inspection of fabric resistance to abrasion by Fourier analysis

A method based on the angular correlation of the Fourier spectra of fabric images is proposed to automatically evaluate web resistance to abrasion.

Unsupervised novelty detection using Gabor filters for defect segmentation in textures

Gabor wavelets are applied to develop an unsupervised novelty method for defect detection and segmentation that is fully automatic and free of any adjustable parameter. The algorithm combines the Gabor analysis of the sample image with a statistical analysis of the wavelet coefficients corresponding to each detail. The statistical distribution of the coefficients corresponding to the defect-free background texture is calculated from the coefficients distribution of the sample under inspection. Once the background texture features are estimated, a threshold is automatically fixed and applied to all the details, whose information is merged into a single binary output image in which the defect appears segmented from the background. The method is applicable to random, nonperiodic, and periodic textures. Since all the information to inspect a sample is obtained from the sample itself, the method is proof against heterogeneities between different samples of the material, in-plane positioning errors, scale variations, and lack of homogeneous illumination. Experimental results are presented. Some results are compared with other unsupervised methods designed for defect segmentation in periodic textures.

Referenceless segmentation of flaws in woven fabrics

The automatic segmentation of flaws in woven fabrics is achieved by applying Fourier analysis to the image of the sample under inspection, without considering any reference image. No prior information about the fabric structure or the defect is required. The algorithm is based on the structural feature extraction of the weave repeat from the Fourier transform of the sample image. These features are used to define a set of multiresolution bandpass filters, adapted to the fabric structure, that operate in the Fourier domain. Inverse Fourier transformation, binarization, and merging of the information obtained at different scales lead to the output image that contains flaws segmented from the fabric background. The whole process is fully automatic and can be implemented either optically or electronically. Experimental results are presented and discussed for a variety of fabrics and defects.

Objective measure of woven fabric’s cover factor by image processing

Digital image processing techniques have been applied to perform an automatic method for the objective measure of woven fabric’s cover factor. Based on a frequency domain analysis, digital images of woven fabric samples, obtained with a camera assembled to a microscope, were cropped to enclose the maximum integer number of warp and weft periods and leveled for non-uniform illumination. Posterior thresholding, designed to perform satisfactorily for both high and low cover factor fabric samples, gave rise to the objective value. The method was applied to three different sets of samples manufactured in plain weave, with known yarn numbers and thread counts. Cover factors obtained by this method showed good correlation with those obtained by a set of visual observers and were consistent with woven fabric parameters: yarn numbers (tex) and thread counts (yarns/cm). The procedure could be useful to monitor mean cover factor as well as cover factor variability in fabric batches. It does not require sophisticated equipment and could be straightforwardly implemented in a textile analysis laboratory.

Fabric inspection by near-infrared machine vision

Fabrics that have superstructures of colored squares, bands, etc., superimposed upon the basic web structure can be advantageously analyzed by use of near-infrared (NIR) illumination and a conventional monochrome camera. The reduction in contrast of the superstructure signal in the NIR image facilitates inspection of the fabrics structure and defect segmentation. Underdetection and misdetection errors are noticeably reduced. This inspection takes advantage of the residual sensitivity of a monochrome camera, which can reach 1000 nm. The light source is an array of NIR LEDs emitting in a band to which the camera is still sensitive. NIR visual inspection can be performed by machines or by humans. In the latter case the observer looks at the NIR image of the fabric displayed on a TV monitor.

Reducing the circadian input from self-luminous devices using hardware filters and software applications

The widespread use of self-luminous devices at nighttime (cell-phones, computers, and tablets) raises some reasonable concerns regarding their effects on human physiology. Light at night is a known circadian disruptor, particularly at short visible wavelengths, and it seems advisable to have practical tools for tailoring the spectral radiance of these displays. We analyse two possible strategies to achieve this goal, using hardware filters or software applications. Overall, software applications seem to offer, at the present time, the best trade-offs for controlling the light spectra emitted by existing devices. We submit that such tools should be included as a standard feature on any self-luminous device and that their default settings should be established according to the best available knowledge on the circadian effects of light.