A comparative study of texture analysis algorithms in textile inspection applications

Abstract

Nowadays, quality control is an important problem for fabric manufacturers. Typically these operations have been carried out by humans operators. However, this method has numerous drawbacks such as low precision, performance and effectiveness. Therefore, automatic inspection systems have increased substantially in the last decade. This work evaluates the performance of some texture measures in textile defect detection applications. For classification a method based on leaving-one-out is used. Our study has been carried out using a large database of samples to take into account a wide spectrum of fabrics and multiple defects of different nature reported by specialized works and publications. A ranking with the effectiveness of best algorithms is presented for every type of fabric. In addition, the computation time of algorithms is compared.