Evaluation of crack propagation in concrete bridges from vehicle-mounted camera images using deep learning and image processing

In Japan, all bridges should be inspected every 5 years. Usually, the inspection has been performed through the visual evaluation of experienced engineers. However, it requires a lot of load and expense. In order to reduce the inspection work, an attempt is made in this paper to develop a new inspection method using deep learning and image processing technologies. While using the photos obtained by vehicle-mounted camera, the damage states of bridges can be evaluated manually, it still requires a lot of time and load. To save the time and load, deep learning, which is a method of artificial intelligence is introduced. For image processing, it is necessary to utilize such pre-processing techniques as binarization of pictures and morphology treatment. To illustrate the applicability of the method developed here, some experiments are conducted by using the photos of running surface of concrete bridges of a monorail took by vehicle-mounted camera.

Automated summary

This paper discusses the current situation and issues of bridge inspection in Japan, and proposes a new inspection method using deep learning and image processing technologies. By utilizing photos taken by vehicle-mounted cameras, the damage states of bridges can be evaluated manually. To save time and load, pre-processing techniques such as deep learning and image processing are used.