Feasibility Analysis of Tamura Features in the Identification of Machined Surface Images Using Machine Learning and Image Processing Techniques

Document Type

Article

Publication Title

Engineering Proceedings

Abstract

In modern manufacturing industries with Industry 4.0 capabilities, the automated identification and classification of machined surfaces based on their texture will play a crucial role. Texture analysis through computer vision, image processing, classification using artificial neural networks (ANN), and various machine learning techniques have been prominent research areas in recent decade. Tamura features are very popular in selecting optimum textural features from an image, especially in the medical domain. These textural features correspond to human visual perception and play a significant role in identifying and shortlisting the best features from the photographs. Despite the popularity of Tamura features in the medical domain, their usage in extracting the features from machined surface photographs is seldom reported. Hence, the present study investigates the feasibility of using Tamura features to classify machined surface images produced using turning, milling, grinding, and shaping operations in manufacturing. Photographs of the surfaces produced are obtained using smartphone cameras. Further, the photographs are preprocessed and divided into sixteen different portions. Then, Tamura features are extracted and are given as input to ANN, support vector machines (SVM), K-Nearest Neighbor (KNN), Decision Tree (DT), and Random Forest (RF). The result shows that each machine learning (ML) algorithm performs differently while classifying the same set of machined surface images. Amongst the ML algorithms considered in the study, RF classified the photographs of surfaces machined using different machining operations with the highest accuracy. On the other hand, SVM performed poorly.

DOI

10.3390/engproc2023059092

Publication Date

1-1-2023

This document is currently not available here.

Share

COinS