A Survey on Various Fabric Defect Detection Methods

Abstract

A survey on various fabric defect detection algorithms is conducted. A local homogeneity, mathematical morphology based novel fabric flaw identification algorithm is implemented. Initial phase includes the construction of a neoteric homogeneity image (H-image), from which the local homogeneity of every pixel is calculated. From the H-image we arrive at the Histogram which is required to select the suitable threshold value which results in producing the Binary image. Using the Binary image we can excerpt the convenient size and shape of the Structuring Element (SE) that is needed for mathematical morphology. In a second phase, a sequence of Morphological operations are carried out on the image with the Structuring Element in order to identify any flaws in the garments. Simulation outputs depict perfect flaw identification having false alarms to be less. Secondly, we put forward an automated image processing for identifying stripe deficiencies in circular shaped knitted cloth materials. We represent how a clearly viewable flaw can be optically embellished to upgrade human verification and how image processing based on descriptor and machine learning could be utilized to permit automatic stripe identification. Finally in this study, data sets obtained by applying local binary pattern and gray level co-occurrence matrix feature extraction methods on Tilda textile data are trained with artificial neural networks and two different models are created and success rates are compared.