2022 Volume 1 Issue 1 Pages 439-444
A bridge inspection needs a lot of costs. It causes a lack of engineers and budget. So some local governments couldn’t complete the bridge’s aggressive preventive maintenance in Japan. We focus on the image-based bridge inspection technique. It enables us to reduce bridge inspection costs. Some previous studies proposed a deep learning-based method for crack detection. These methods detect cracks on a concrete surface image of the bridge and show high accuracy. However, the evaluation of the robustness of such a method to the real bridge surface images is not enough yet.
In this paper, we proposed a multi-model method to detect cracks on various bridges. This method uses several models trained with different fake images. We evaluated three different crack detection models with three different bridge images. The evaluation result shows the effectiveness of multi-model trained by different fake images.
