Applying a co-occurrence matrix to automatic inspection of weaving density for woven fabrics

This paper evaluates the efficiency and accuracy of a way to detect a fabric's weaving density using a co-occurrence-based method. Three basic fabric weave structures-plain, twill, and satin-are evaluated by this method. Based on the co-occurrence matrix algorithm, the weaving density of a plain weave structure can be computed exactly. The method used to process image feature extraction, the co-occurrence-based method, is one in which a feature parameter (the contrast parameter or CON) is obtained. It consists of contrast measurements involving sixty-four spatial displacements (i.e., 1-64) and two directions (0 and 90 degrees) of fabric images used for calculation. The results show that the calculation precision for the plain weave is far better than that for the twill and satin weaves.