2020 Volume 72 Issue 6 Pages 3-7
It is useful to utilize paper-based archive data of various kinds, in order to analyze the low frequency phenomena such as debris flow disaster by volcanic activity. SfM-MVS (Structure from Motion-Multiple View Stereo), which is able to create a 3 D shape by means of multiple-view photos, has been increasingly used in the field of geomorphology and disaster mitigation. In this study, we reconstruct Digital Surface Model (DSM) through SfM-MVS by using the archived time-series paper-based aerial photographs of volcanic activity of Mt Unzen from 1990. In our results, the DSM difference of more than about 6 m is evaluated to be significant by error analysis. In order to decrease these error, though it is difficult to match the same Ground Control Points (GCPs), which are necessary to transform the geographic coordinate system, on these archived photographs between different periods because of topographic change etc., it is useful to set the GCPs by using more accurate location such as LiDAR data or different idea for minimizing error propagation.
