Identification of Efflorescence in Reinforced Concrete Slab of Road Bridge by Decision Tree

Abstract

In recent years, deterioration and damage of concrete structures have become obvious due to long-term use. It is necessary to detect the deterioration and damages easily and accurately. Identifying efflorescence is an important issue for maintaining RC slabs. In this study, we attempted to identify efflorescence using a decision tree, which is one of machine learning techniques, in reinforced concrete slabs of road bridge that constructed 50 years ago. Two feature values were set as explanatory variables: luminance value and pixel value after DoG (Difference of Gaussian) filter. As a result, the proposed method was higher than the Otsu method, which is one of discriminant analysis methods, in all indicators of accuracy, recall, precision, and F value. The three indicators of accuracy, recall, and F value were 0.8 or higher. These results suggest that the proposed method is useful for identifying efflorescence.