Automated crack pattern recognition from images for condition assessment of concrete structures

Abstract Evaluating the appearance of cracks on concrete elements is necessary during condition assessment. Most imaging techniques related to crack identification focus on detecting cracks and measuring basic dimensions of cracks. However, condition assessment of an element from visible cracks requires further information, of which characterizing the crack pattern is necessary. To achieve this, this paper proposes an automated approach to recognize concrete crack patterns from images. By analyzing the characteristics of structural and non-structural cracks, a binary classification of crack patterns into isolated patterns and map patterns is proposed. The recognition of crack patterns is performed through similarity comparisons using DISTS index. Various parameters that may affect the performance are investigated through several experiments conducted using real-world images. According to the experimental results, the proposed method is shown to successfully recognize crack patterns with an accuracy of over 96%. Recommendations are further proposed to enhance the performance of the method.