Measurement of Iron Rust Area Ratio by Image Processing Using Color Photography and Machine Learning

Abstract

A method using binarized image data is common to calculate the area of iron rust in a digital image system. However, iron rust includes red rust as well as black corrosion products, and it may be inappropriate to use binarized data. Therefore, we tried to evaluate with a 24-bit color image of sRGB. The areas of red rust, black corrosion products, and metal substrate of a small area iron rust image were visually judged, teacher data was created, and evaluation was performed using a support vector machine. As a result, the image was divided into each area, and the area and its ratio could be calculated.