東日本大震災の被災と復興の画像アーカイブの構築とコンピュータビジョンによる被災市街地の時空間モデリング

抄録

From April of 2011, we have been continuing to obtain video records of 3.11 Large Earthquake and Tsunami disasters at Tohoku coastal areas in Japan. We install 360 °directional camera on a special car, and are visiting same areas at every one or two months, and obtained detail images at every 1m to 2m step on streets. Total amount of image data up to now, for six year activity, becomes more than 150 million 360 °scenes of 100 Tera Bytes. We construct a video archive for recording the real detail damages of towns, industrial areas, and agricultural lands. The main purpose of this activity is to visualize the process of reconstruction and long term recovery from the disaster. By those image sets, employing the computer vision techniques, we have been studying the spatial modelling of the temporal changes of city structure by the disaster and the afterward recovery process. They are summarized in this paper. One of additional role of our detail spatial and temporal observation on surface is to bridge between wide and global satellite or aerial observations and real local human lives. The combination of those observations with different spatial and temporal resolution and different view-points is important to estimate the effectiveness of the recovery from the serious disaster.