UAVによる樽前山溶岩ドームの３次元地形モデリング

概要

Monitoring of volcanic activity is essential for volcanological studies as well as disaster mitigation. In case of a lava dome eruption, it is particularly important to monitor the instability of the dome that may result in its collapse accompanied by a pyroclastic flow. For a study of the growth mechanism of the lava dome and the hazard assessment against the dome collapse, frequent monitoring of the topography with a high resolution is necessary. The airborne photogrammetry and/or the space geodesic technologies are commonly used to observe the ground surface. However, they are usually not suitable for frequent operations from the viewpoint of its cost and/or the recurrence period. Therefore, in this study, we tried the UAV (Unmanned Aerial Vehicles) photogrammetry for the lava dome topography measurement. This study aimed for evaluating the accuracy of the topography reconstruction by the drone-based photogrammetry as a preparatory step for the future monitoring on a regular basis. We chose the lava dome of Mt. Tarumae as a test field, and tried reconstructing a 3D topography model consisting of dense point cloud and the DSM (Digital Surface Model) from hundreds of aerial images taken by a drone. We succeeded in reconstructing the topography models with or without the GCPs (Ground Control Points), and finally obtained the DSM of a high spatial resolution about 9.5 cm/pixel. To evaluate the accuracy of our DSM, we compared it with the DEM (Digital Elevation Model) provided by the Geographical Survey Institute to estimate the elevation difference between them. Contrary to our expectation, the DSM with the GCPs had a larger self-inconsistency than the one without the GCPs. We suspect that spatially imbalance arrangement of the GCPs may have affected. Our achievement of generating the high-resolution topography model demonstrated that the drone-based photogrammetry is useful for volcano observation. The technique will be more common and easier with improving performance of UAVs in the future.

この論文で使われている画像