Abstract
In many developed countries with a long history of urbanization, there is an increasing need for automated computer vision (CV)-based inspection to replace conventional labor-intensive visual inspection. This paper proposes a technique for the automated detection of multiple concrete damage based on a state-of-the-art deep learning framework, Mask R-CNN, developed for instance segmentation. The structure of Mask R-CNN, which consists of three stages (region proposal, classification, and segmentation) is optimized for multiple concrete damage detection. The optimized Mask R-CNN is trained with 765 concrete images including cracks, efflorescence, rebar exposure, and spalling. The performance of the trained Mask R-CNN is evaluated with 25 actual test images containing damage as well as environmental objects. Two types of metrics are proposed to measure localization and segmentation performance. On average, 90.41% precision and 90.81% recall are achieved for localization and 87.24% precision and 87.58% recall for segmentation, which indicates the excellent field applicability of the trained Mask R-CNN. This paper also qualitatively discusses the test results by explaining that the architecture of Mask R-CNN that is optimized for general object detection purposes, can be modified to detect long and slender shapes of cracks, rebar exposure, and efflorescence in further research.
Highlights
Most of highly developed countries have difficulties in managing an increasing number of old civil structures
Before applying the trained model, all test images were resized to 1024 × 1024, which is the input size of the Mask R-convolutional neural network (CNN) model, regardless of their own resolutions
The overlapping pixels are categorized as true positives (TPs), the pixels of resulting masks that do not overlap with the ground truths are categorized as false positives (FPs), and the pixels of ground truths that do not overlap with the resulting masks are categorized as false negatives (FNs)
Summary
Most of highly developed countries have difficulties in managing an increasing number of old civil structures. This study proposes a concrete damage detection method using an instance segmentation deep learning model called Mask R-CNN (Mask and Region-based Convolutional Neural Network) [33]. The scope of this study is on detecting multiple types of concrete damages and does not include the quantification of detected damages, which has already been proposed in the previous paper by the authors [17] This method alters conventional visual inspection in the field environment with surrounding objects. Beyond the detection of concrete cracks [17], spalling, rebar exposure, and efflorescence are selected as types of damage to be detected by the Mask R-CNN model to alter the conventional visual inspection of concrete structures.
Sign-in/Register to view highlights & summary Sign-in/Register to access full text options 
Free) 
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a call
