Rust classification using image analysis of steel structures

The purpose of this paper is to classify the rust conditions of steel structures by using image analysis. For classification purposes, support vector machine (SVM) was utilized. The purpose of our research is to provide additional information on rust conditions for reconfirmation use, prevent errors during rust evaluations, and to create a robust and practical rust evaluation system. For our methodology, we took photographs of steel structures using a digital camera and divided those images into smaller regions in order to calculate the evaluation parameters. The parameters themselves were determined by using a rust recognition process. We then used the classification results provided by evaluation experts and the parameters as SVM input vectors in order to classify rust conditions. As a result of our efforts, we developed an evaluation system with a correct classification rate of 99% for learning sets and a correct classification rate of 66% for test sets. The results we obtained suggest that the creation of a rust evaluation system using this method is possible.