Tunnel defect detection: Segmentation approaches

With the industry 4.0 paradigm and the growth of economies worldwide, various infrastructures such as tunnels, bridges and skyscrapers have been constructed, which are indispensable for our daily life and are used by large number of people in their daily lives. Therefore, infrastructures built more than five decades ago are experiencing aging problems. Every year 4700 km of new tunnels are built, resulting in annual growth of 7%. Accordingly, the total amount of tunnels that must be inspected will increase in the future. However, traditional tunnel surface defect inspection mainly relies on naked eye inspection, which is carried out by the technical inspectors walking along the tunnel line. Thereafter, this method is inefficient and costly in examining large-area tunnel lining structures. In the following years, the use of deep learning has shown outstanding performances in many industrial fields such as object detection semantic segmentation and instance segmentation. This thesis is focused on analysing some of the state-of-the-art computer vision models for segmentation and object detection with aim of detecting humidity defects in Tunnels. The experiments are heavily dependent over deep learning models that have proved to best in the field of computer vision in the past. Moreover, economically this approach is need of the hour and in line with the requirement of Industry 4.0 revolution. Finally, successful implication and deployment of this technique will not only reduce overall labour cost and time but will also turn out to be beneficial for durability of aging and defected tunnels.