| description abstract | Moisture marks (wet areas) are significant defects that may develop on the surfaces of subway structures as a result of water leakage through soil. The detection and assessment of moisture marks are predominantly conducted on the basis of visual inspection (VI) methods, which are known to be costly, labor-intensive, and error-prone. The objective of this paper is to develop an integrated model based on image processing techniques and artificial intelligence to automate consistent moisture marks detection and numerical representation of the distress in subway networks. The integrated model comprises sequential processors that automatically detect moisture marks on concrete surfaces, and artificial neural networks (ANNs) for moisture marks identification and quantification. First, red-green-blue (RGB) images are preprocessed by means of spatial domain filters to denoise the image and enhance the crucial clues associated with moisture marks. Second, a moisture detector is streamlined with a set of morphological algorithms to detect wet areas. Third, the area percentage and severity of moisture marks are measured using the ANN model in conjunction with cross-entropy optimization function. The integrated model was validated through 165 images. Regarding the moisture marks detection algorithm, the recall, precision, and accuracy attained were 93.2, 96.1, and 91.5%, respectively. The mean and standard deviation of error percentage in moisture marks region extraction were 12.2 and 7.9%, respectively. Also, the ANN model was able to satisfactorily quantify the moisture marks area with an average validity of 96%. The integrated model is a decision support tool, expected to assist infrastructure managers and civil engineers in their future plans and decision making. | |
