earticle

영어

Digital photographs taken with the microscope are practically used to diagnose deteriorated membrane materials. This paper presents a new automated diagnostic model utilizing image analysis instead of checking with human eyes. We first extract from a digital image statistics of color information and features of the objects that emerge along with deterioration such as cracks on the surface of polyvinyl chloride coating and exposed fabrics. Then these features are used as explanatory variables for building prediction models which are constructed based on multiple linear regression and M5P method which is combination of multiple linear regression and decision tree. As the result of the experimental tests, our prediction models exhibited more than 90% accuracy in coefficient of correlation.