Model Development of Marble Quality Identification Using Thresholding, Sobel Edge Detection and Gabor Filter in a Mobile Platform

Abstract

Organizations in the marble industry use machines to identify the marble slab’s quality which is costly for developing countries like the Philippines. Marble is classified by expert human visually and classification is prone to error. This paper presents a study on marble classification using image processing based on color and texture. Features extraction using Thresholding, Gabor Filtering, Sobel Edge and Local Binary Pattern (LBP). Three supervised learning was used which includes Support Vector Machine, Decision Tree, and Random Forest. 120 marble images were trained and 75 images were used for testing. LBP is consistent with 86.67% accuracy, 0.800 of Kappa and execution-time of 2 seconds using the Decision tree. The Model was applied to the prototype with 82% accuracy to 100 unlabeled images tested with the expert. In conclusion, the developed model can classify the marble quality which is higher than the accuracy from the previous research work applied to the industrial machines.