Application of ‘vision in the loop’ for inspection of lace fabric

https://doi.org/10.1016/S1077-2014(98)90002-XGet rights and content

Abstract

Within the textile manufacturing environment, inspection is an important process which is employed in order to ensure high quality of the final products. Traditionally, this process is achieved manually which is very tedious, time consuming and labour intensive. A central problem in automatic visual inspection and computer vision is to determine the extent to which one shape differs from another. This is the key element of any inspection algorithm. Pattern recognition operations such as correlation, template matching and model based vision methods can all be viewed as techniques for determining the difference between shapes. In this paper the problem of visual inspection of deformable materials in general, and lace in particular, is considered. This is particularly difficult to achieve due to inherent and inevitable variations between a ‘model’ and the material being inspected. A mechatronic approach based on correlation along with morphological filters and ‘active vision in the loop’ using a line-scan CCD camera is presented. The utilized algorithm and its advantages and disadvantages are also discussed.

References (21)

There are more references available in the full text version of this article.

Cited by (27)

  • Integral images-based approach for fabric defect detection

    2022, Optics and Laser Technology
    Citation Excerpt :

    To reduce computation time, they carried out the difference operation in the wavelet-domain instead of doing the subtraction in binary images mode. Yazdi, and King [21], Shankar and Zhong [26] used image difference method for defect detection on lace fabric and wafer surface, respectively. Their results showed that images difference method was very sensitive to noise and to shift variations, consequently there were many false alarms.

  • Fabric defect inspection based on lattice segmentation and Gabor filtering

    2017, Neurocomputing
    Citation Excerpt :

    Sub-band selection for reconstruction is based on the analyses of the wavelet coefficient energies, and the defects are identified by thresholding the reconstructed images. AR-based method in [16] arranges pixels of an image to a single row and scans the pixel sequentially, the gray value of pixel under scanning is compared with the ideal value predicted by the 1D autoregressive model w. r. t. the previously-scanned pixels. The model parameters are trained by minimizing the prediction error for defect-free samples.

  • A timely detection of a coated board streak defect in subsampling conditions using monochrome vision system

    2012, AEU - International Journal of Electronics and Communications
    Citation Excerpt :

    Since the installed system uses 2D cameras it satisfies the primary requirement for detecting streaks. Systems capable of detecting a 1 mm × 1 mm size defect typically operate at a 1.5 oversampling rate [7]. Since the vision system resolution is 1 pixel/mm of coated board, this means it would operate in the subsampling condition, since Nyquist sampling theorem is not satisfied as defined in [3].

  • Automated fabric defect detection-A review

    2011, Image and Vision Computing
    Citation Excerpt :

    Section 7 addresses the future direction and conclusions. As observed from real implementation of fabric inspection systems, it is necessary to consider factors such as (1) contrast between defects and texture surface [20,37]: e.g., a low contrast in image easily leads to a misclassification; (2) consistency of texture background [37]: e.g., color difference and distortion along a texture affect image acquisition; (3) resolution of input image [19,37–41]: e.g., a low resolution image cannot show fine defects in fabric; (4) alignment of input image [3,42,43]: e.g., misalignment in image acquisition induces false defect detection in template matching approach; (5) size [37,44], and shape [45–47] of defects: e.g., defect of small size or defect similar to a pattern shape increases difficulties in recognition; (6) speed or computation complexity of defect detection [37,48,49]: e.g., long learning delays may not be practical; (7) lighting [13]: e.g., improper illumination yields poor resolution and contrast; and (8) image acquisition techniques: e.g., most inspection methods use digital cameras to capture images. However, alternative approaches are also available, such as near-infrared (NIR) [50], X-ray, multispectral imaging and polarimetry, which may provide extra features in defect detection.

View all citing articles on Scopus
View full text