A novel multiscale-multidirectional autocorrelation approach for defect detection in homogeneous flat surfaces

Abstract

Defect detection in flat web surface products is a challenging task. Reliable vision-based systems for detection of defects require the suitable selection of a huge set of parameters which highly impact the performance of these systems such as image resolution/scale, size of the scanning window, feature extraction, direction of scanning, classifier type and parameters and system performance evaluation measures. This paper addresses these issues and introduces a novel multi-scale and multi-directional (MSMD) autocorrelation function (ACF)-based approach for reliable defect detection and localization in homogeneous web surfaces. The proposed approach has been experimentally tested on samples from the well-known TILDA textiles database and wallboards. Performance evaluation using the system Precision, Recall (Sensitivity), Specificity, Accuracy, Youden's index, F-measure and Matthews correlation coefficient has shown that the MSMD ACF approach outperforms the state-of-theart approaches like MSMD Log-Gabor filters. The MSMD ACFs approach results in better performance indicators for defect detection than the Log-Gabor based approach in addition to being about 2-6 times faster in defect detection.

Homoscedasticity for defect detection in homogeneous flat surface products

“…This study has been conducted to implement an efficient defect detection technique that achieves higher accuracy compared with that achieved by the state-of-the-art approaches such as LG wavelets 3 and MSMD ACF 4 but with lower computational complexity. A performance comparison (Tables 10-12) of the proposed LT approach with the MSMD approaches discussed by Tolba 4 indicates the promising performance of the LT approach. Comparison of LT with local binary patterns (LBP) 39 shows that LT outperforms LBP approach in defect-detection accuracy while LBP needs 64% of the time needed by LT for feature extraction.…”

“…Comparison results Comparison of LT detection with the results of difficult cases reported by Tolba et al3,4 …”

“…The major contributions of this work are: a novel application of LT for defect detection in industrial flat surface products, and a fast and efficient context based algorithm for defect detection compared with the stateof-the-art approaches such as log-Gabor (LG) filters, 3 multiscale multidirectional autocorrelation, 4 Gabor wavelets, [5][6][7][8][9][10][11][12] wavelet transform-based techniques, 13 and filter-based approaches. 14,15 The implemented defect detection approach achieves better results compared with those achieved by the state-of-the-art approaches such as LG wavelets but with lower computational complexity resulting in a 6-10 times faster defect detection system.…”

The Use of Haar Wavelets in Detecting and Localizing Texture Defects

“…For the same purpose, an optimized elliptical Gabor filter is considered (Hu, 2015). Tolba (2012) introduces a novel multi-scale and multi-directional (MSMD) autocorrelation function (ACF)-based approach to reliable defect detection and localization on homogeneous web surfaces.…”