Local binary pattern based marble classification application using intelligent pattern recognition system using the extreme learning approach

Separation of the marbles according to the pattern quality is an operation made according to expert decision. The marble classification stage is the most critical part of the economic value. In this study, a system was proposed which performs the classification of marbles quickly and with high performance and learns in doing so. This system applies feature subtraction to the images obtained from the camera or memory. The Local Binary Pattern (LBP) feature-based Extreme Learning Machine (ELM) was used to classify marble views. The training set is obtained by applying LBP feature extraction to the image set. The ELM classifier performs learning using the LBP-based training set. The Ensemble classifier was used to measure the classification success of the ELM classifier. C4.5 based decision tree, forward feed static advisory artificial neural network and support vector machine classifiers constitute collective classifier. The classification performance of the ELM is compared with the collective classifier. Multiple feature extraction was used to compare the performance performance of the LBP-based training set. Histogram and Scale Invariant FeatureTransform (SIFT) based attributes are available for multiple feature extraction. The Bag-of-Words (BOW) model was used to transform SIFT-based features into a training set. The k-means clustering method is used to distinguish visual words in the vocabulary. Graphical interface (Graphical User Interface-GUI) is used to develop the feature extraction and classification processes of the prepared models. By training the system through the application, the user can determine the class of marble that does not know the class. The GUI developed within the scope of the study is suitable for industrial use.