Abstract:
Structural Health Monitoring (SHM) refers to a periodical inspection to monitor the
condition and characteristics of civil structures. On the other hand, damage can be
interpreted as a structural modification that changes physical properties and weakens the
structure, so it should be addressed as early as possible to prevent additional damage.
Usually, cracks are visually monitored by inspectors who record data regarding presence,
location, and width. Manual visual inspection is often deemed ineffective in terms of
cost, protection, the accuracy of assessment, and reliability. As moving with the fast pace
of technology advancements, the possibility of the information technology-driven
methodologies in the construction field is also getting wide visibility. In surface crack,
detection different technology-backed automated systems outperform the traditional
manual inspection and crack detection. With the help of different computational aids like
Image Processing, Machine Learning, and Deep Learning techniques, the images, and
videos captured from surveillance sites are analyzed for automated crack detection.
In this study design and development of a concrete surface crack analysis system was
done using image processing machine learning and deep learning techniques. The major
components include an acquisition module, an analysis module, and a client application
module. The acquisition module is designed and developed as an IoT device that can be
controlled over the internet using a mobile application, and it is intended to move over
the surface and capture the cracks. The analysis module is integrated with the web server,
build using python programming and flask library. The core analysis process was
implemented using computer vision and deep learning algorithms. The classification of
crack images to cracked or non-crack was done by convolutional neural networks, the
segmentation and localization are carried out with image processing. The client
applications have two versions, one is a web app and another is a mobile app that is used
to control the accusation and to interact with analysis. The system is validated with a set
of crack and non-crack images of beams and walls collected from the concrete
technology lab, TKM.
