ABSTRACT
In this paper, we visually deciphered the damage to road structures in 25 aerial photographs which were taken after the 2016 Kumamoto earthquake and detected these damage modes according to the image training database of seven damage modes: road surface cave-in, road surface liquefaction, road surface cracks, road surface collapse, road surface steps, falling rocks and landslides and building debris on road surface. Specifically, texture analysis produced a histogram of six feature values: contrast, variance, skewness, kurtosis, energy and entropy, and a threshold were set to detect damage. The results presented that the coefficients of variation of kurtosis and entropy were small and effective for damage detection.
