Automated steel defect detection using image processing and machine learning

Abstract

Automated Steel Defect Detection application provides the user with the comprehensive defect detection and analysis of the steel surface. The system provides the defect detection of a steel surface by visualizing and processing the defect region in the image and extracts features from that region. The feature set is used to predict the defect class using the Support Vector Machine (SVM) classifier. The system resolves the issue of extraneous labor and provides an analysis of the defect in just a few clicks. The system is user friendly and is relatively simple in terms of usability. Users can track the progress of the operations performed by analyzing the status bar provided in the graphical user interface. The system is a desktop application and can be integrated into any environment since the application does not require additional system resources to operate. The application provides the solution to the problem of manual inspection and is a step towards the advancement of the steel industry. Rigorous testing has been conducted on the unit level as well as the system level to improve the performance and usability of the system. The detection of defects in a steel surface will help the analyst to make decision based on defect severity. The application makes this process simple and visually pleasing.