DESIGN AND DEVELOPMENT OF CONCRETE SURFACE CRACK ANALYSIS SYSTEM

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.