Study on image diagnosis of timber houses damaged by earthquake using deep learning

The image processing (with deep learning, semantic segmentation, hough transformation, edge extraction) for the image diagnosis of wooden houses damaged by earthquake was studied and verified. As a result, the following 3 aspects were revealed. 1. The usefulness of chroma keying for creating image database for deep learning instead of real damaged images 2. The potential of the image diagnosis for not only qualitative damage assessment of timber houses, but also quantitative one 3. The possibility of installation of drift ratio estimation into the image diagnosis.