Defect Detection of Tiles Based On High Frequency Distortion

Abstract

Quality control in Tiles Industry is of great importance. Therefore, it is effective to improve an automatic inspection system, instead of manpower, to increase accuracy and velocity and decrease costs. To this end, a new method to segment tile surfaces is offered in this study. This method aims at detecting defective areas in a tile, based on extracting features of edge defects. This method is based on the idea that human eye can better perceive the defects in a tile by looking at its edges. In the proposed method, first, in order to extract frequency characteristics resistant against transference, Undecimated Discrete Wavelet Packets transform is applied on images. Later, by computing local entropy values on high-frequency sub-bands images, those which appropriately include images defects are chosen to extract statistical features. Finally, Back propagation neural network method is used to determine segmented images containing defective areas. The obtained results, both visually and computationally indicates the higher efficacy of this method compared with the related state of the art methods. DOI: http://dx.doi.org/10.11591/ijece.v3i4.3014