Automated classification of hand-woven and machine-woven carpets based on morphological features using machine learning algorithms

As a cultural heritage, hand-woven carpets engender trust and admiration in individuals who recognize their authenticity. It is the expertise of experts who determine whether a carpet is hand-woven or machine-woven based on authenticity criteria. A total of 48 morphological features were extracted by three carpet experts from 359 handwoven and machine woven carpets. Machine-learning algorithms were used to classify the extracted features. With an accuracy of 96.66%, the ANN algorithm achieved the best classification performance. Afterward, 28 morphological features were selected with the highest gain ratios and reclassified. Based on the selected features, SVM (Support Vector Machine) achieved the best classification accuracy of 96.66%. A carpet expert performed classification using the morphological features extracted by machine learning algorithms to evaluate the classification results obtained through machine learning algorithms. When 48 morphological features were used, the accuracy was 98.61%, and when 28 morphological features were used, the accuracy was 97.77%. Based on the results, artificial intelligence techniques are suitable for detecting morphological features in hand-woven and machine-woven carpets and for automatically classifying them.