Detection of Texture Defects in Tiles using Imaging Techniques

Abstract

Tile Inspection is a tedious process in manufacturing factories. Quality control in manufacturing tiles is hard, labor intensive and it is performed in a harsh industrial environment with noise, extreme temperature and humidity.  Since the visual appearance of the material and other biological products is a major factor in the judgment of quality, classification is an important part of quality control in the manufacturing industries. This classification has historically been performed by use of the only “tool” available, the human eye. In this paper, we present a image processing techniques which detects defects in tiles.

The method uses a segmentation algorithm in which the color images of tile samples are grouped into similar color characteristics.

Before the segmentation, the application of Gaussian Low pass filter becomes necessary for the image to blur sharp edges. After that, each part is individually analyzed with its characteristics of color and compared against a given appropriate profile for the particular tile sample. Due to the false background of pixels, the algorithm was able to establish the amount of color changes in a given sample with minuteness and with least error. The segmentation and rejection process is tested with several different tile samples and the results are compared with visual inspection results. The results of the comparison were quite good.