Thank you for Subscribing to Utilities Business Review Weekly Brief
Utilities Business Review | Friday, February 03, 2023
The underground infrastructure of cities plays a vital role in providing essential services to their residents. Researchers are exploring various construction methods to facilitate underground infrastructure projects' activities by integrating digital technologies.
FREMONT, CA: An important part of the underground infrastructure lifecycle is operation and maintenance, which involves many procedures and activities to ensure the infrastructure's normal functioning, including regular inspections and monitoring, condition assessment, and maintenance planning. Contractors can improve efficiency in construction activities through different digital technologies and methods, including automation and other digital solutions.
Current research on the operation and maintenance of underground infrastructure mainly focuses on the following:
Stay ahead of the industry with exclusive feature stories on the top companies, expert insights and the latest news delivered straight to your inbox. Subscribe today.
Automated inspection data interpretation: A typical inspection of underground infrastructure involves using various devices, including visual inspection devices like CCTV and SSET, electromagnetic and radio frequency devices, and acoustic devices. The inspectors must manually identify defects during the inspection and review the collected data afterward to assess the infrastructure's condition based on their experience and expertise. Despite requiring enormous time and effort, traditional inspection practices still suffer from human error. Various methods have been developed to identify the defect's existence, type, or location from inspection data like CCTV videos, GPR images, and acoustic signals. As part of the interpretation, infrastructure components, such as joints and taps of pipes, are identified. According to the characteristics of the data obtained, different types of interpretation methods have been proposed.
Computer vision: The dominant forms of computer vision, such as conventional image processing techniques and deep learning-based approaches, are used to interpret visual inspection data automatically. Traditional computer vision workflows involve multiple stages, including preprocessing, feature extraction, classification, and detection of defects using machine learning algorithms. It is common to extract features from images by using traditional image processing methods, such as morphological operations, edge detection, and thresholding. A number of studies have used morphological methods alone or in combination with other techniques, such as edge detection, for the extraction of features. Machine learning classifiers, such as neural networks, support vector machines (SVM), and random forests (RF), use the extracted features to detect and classify defects. Traditional workflows require manually constructing features and adjusting classifier parameters for each case.
Deep learning: A deep learning-based approach is more straightforward because it trains a model using annotated data and then applies it to produce results. It is possible to classify defects, detect them, and segment them for underground infrastructure using CNN-based models. A classification process identifies defects in an image or predicts the type of defects for the whole image. In contrast, a detection process identifies defects and localizes their location with bounding boxes. The semantic segmentation task consists of predicting the labels of pixels on a pixel-by-pixel basis. The analysis results are useful for identifying the areas of the image occupied by different defects.
More in News