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.
