Search

The underground sewer network is a vital public infrastructure in charge of large-scale wastewater collection and treatment. Complex defects can occur in sewer pipes due to various internal and external factors, which increase the demand for frequent inspection. Previous defect detection research mainly depended on manual inspection, which is tedious, costly, and error-prone. This study suggests an automatic pixel-level sewer defect segmentation framework based on DeepLabV3+, which can recognize the defect's type, location, geometric information and severity. The impacts of various backbones and pre-processing methods on the model's performance were carefully evaluated. In addition, four state-of-the-art segmentation models (U-Net, SegNet, PSPNet, and FCN) were compared with the presented model to demonstrate its superiority. The experimental results revealed that the DeepLabV3+ with the Resnet-152 backbone structure efficiently identified ten defect types under challenging conditions. The obtained mean pixel accuracy and mean intersection over union (IoU) were 0.97 and 0.68, respectively. In terms of severity analysis, it was revealed that the framework outputs were consistent with the NASSCO pipeline assessment certification program (PACP). In addition, during the testing process, the proposed frame reduction algorithm only required about 16% of the original time required to process an input video. Finally, with a generated detailed report for an inspection video, the suggested framework can offer a decision-making base for more precise and efficient defect inspection and maintenance. • A manually collected sewer defect segmentation dataset that contains over 11,124 images. • An efficient sewer defect segmentation framework based on DeepLabv3+. • A frame reduction is introduced to reduce the computational complexity. • Automatic report generation module to support real-life applications. • Defect severity analysis based on the NASSCO PACP program. [ABSTRACT FROM AUTHOR]

[Display omitted] • A manually collected masonry crack segmentation dataset that contains approximately 2670 samples. • An efficient DeepLabV3-based masonry crack segmentation framework. • Actual length measurement of the segmented cracks based on brick unit's size. • The proposed model outperformed previous state-of-the-art crack segmentation models. While there have been a considerable number of studies on computer vision (CV)-based crack detection on concrete/asphalt public facilities, such as sewers and tunnels, masonry-related structures have received less attention. This research seeks to implement an automated crack segmentation and a real-life crack length measurement of masonry walls using CV techniques and deep learning. The main contributions include (1) a large dataset of manually labelled images about various types of Korea masonry walls; (2) a careful performance evaluation of various deep learning-based crack segmentation models, including U-Net, DeepLabV3+, and FPN; and (3) a novel algorithm to extract real-life crack length measurement by detecting the brick units. The experimental results showed that deep learning-based masonry crack segmentation performed significantly better than previous approaches and could provide a real-life crack measurement. Therefore, it has a huge potential for motivating masonry-based structure investigation. [ABSTRACT FROM AUTHOR]

The sanitary sewer is a crucial underground infrastructure of any country that collects wastewater and carries it to the treatment plant. The damage triggered by various factors, such as external interference, long-term corrosion, and uneven distribution of pressure, could lead to various types of defects inside the sewer pipe. Previous studies primarily relied on human visual perception to evaluate the sewage system, which was tedious, time-consuming, and costly. As a result, an efficient and robust sewer defect localization framework was proposed in this manuscript. The main contributions include (1) a novel sewer defect detection system motivated by the state-of-the-art detection transformer (DETR) architecture, which views object localization as a set prediction topic; (2) a defect severity analysis approach based on the transformer's self-attention operation to analyze defect zone of influence and defect grade; and (3) a manually validated sewer defect localization dataset that contains 10 types of commonly appeared sewer defects. The experimental results suggested that the proposed system outperformed the previous standard object detection approaches with the highest mean Average Precision (mAP) of 60.2% on the collected dataset. • A manually collected sewer defect detection dataset that contains over 47,000 images. • An efficient transformer-based sewer defect detection framework. • Severity analysis of the localized defect using the transformer's self-attention weights. • The proposed model outperformed previous state-of-the-art object detection models. [ABSTRACT FROM AUTHOR]