Automatic Ceramic Tiles Defect Recognition

Abstract

The global market for ceramic tiles industry is highly competitive nowadays. Quality control in production process in ceramic tiles industry has been a key factor for retaining existence in such competitive market. Machine vision based ceramic tiles inspection systems are very useful in this respect, because manual inspection is time consuming and not accurate enough. Hence,machine vision based ceramic tiles inspection systems have been drawing plenty of attention of the researches of different countries in order to replace manual inspection. Two difficult problems are mainly posed by machine vision based ceramic tiles inspection systems. They are defect detection and defect detection classification. Even though there has been plenty of research addressing the defect detection problem, the research aiming at solving the classification problem is scarce. Moreover, scene analysis and feature selection play a very important role in defect detection process. If scene analysis is not properly done, a weak and inappropriate set of features will be selected. Selection of an inappropriate feature set makes the subsequent steps complicated and the classification task becomes harder. In this thesis, a possible appropriate set of features are confined to the spatial domain. We justify the features from viewpoint of discriminatory quality and defect difficulty. The features are extracted using a statistical defect extraction process. We perform defect detection on the applicability of convolutional neural network models by CV2 in the context of tiles defect. We observe the effect of tuning different network parameters and try to explain the reasons. We empirically find four types of defected tiles, including multi label defected tiles with OpenCV using mask image .Finally, we use accuracy metrics to evaluate the models and find out the accuracy of defected tiles detection.