Sliding Windows: A Software Method Suitable for Real-Time Inspection of Textile Surfaces

Abstract

There is a strong interest in expert systems to assist in complex defect identification procedures. To increase accuracy, attempts are made to enhance traditional human inspection by automated visual systems, which employ cameras and image processing routines. It is important that the software core of such a system is based on a robust, fast texture segmentation technique. Many texture segmentation techniques have been pre sented in the literature. Unfortunately, despite the fact that they achieve good results, they are computationally complex and therefore not suitable for real-time applications, since as the complexity of an algorithm grows, it becomes more and more difficult to execute the image examination in real time. This paper presents a detailed description of a fast algorithm for defect detection in textile surfaces and an evaluation of the experimental results for 137 digitized images of representative defects in woven fabrics collected from a larger sample provided by textile companies.