SfMを用いた三次元モデルの生成と災害調査への活用可能性に関する研究

Abstract

画像処理をベースとした形状復元技術および撮影位置推定技術であるSfM を災害情報収集と調査観測に活用するための試行的な研究を行った．小型UAV と手持ちカメラで写真画像を撮影しSfM で三次元モデルを作成した．その結果，斜面災害発生地域において，0.5 m 間隔の細密地形図や位置誤差1 m程度のオルソフォトを生成することができた．また歴史災害資料である1927 年に建立された石碑では，その文字を明瞭に復元，判読することができた．SfM と小型UAV によって，安全，簡便，低コストに，数cm 級の精度を持つ三次元地形モデルの生成や，歴史災害資料の判読ができることを示した．今後，これらの技術は自然災害や地形調査，歴史災害の資料調査等において革新的なツールとして活用されると考える．

This paper reports the results of a trial study for utilizing structure from motion (SfM) with photographs taken from small unmanned aerial vehicles (sUAVs) or handheld camera for disaster prevention research. Low-cost software has also been integrated into the form reconstruction technology, and camera position estimations based on SfM or scale-invariant feature transform (SIFT) image processing can generate three-dimensional models from photographs. Improvements in sUAVs such as the electric multi-rotor radio control helicopter and compact digital camera photography from a low altitude have resulted in lower costs, improved safety, and greater ease of use. We integrated a sUAV with SfM for application to natural disaster sites and historic disaster materials. In slope disaster areas, we generated orthophotographs with errors of less than 1 m and a topographical map with 0.5 m interval contours. Moreover, with respect to historic disaster materials, we were able to restore and decipher very clearly the character of an old and weathered stone monument built after the 1923 Great Kanto Earthquake, where interpretation of the character had previously been difficult. We also developed the necessary expertise to apply this technology. The SfM and sUAVs offer an alternative for generating a highly precise three-dimensional model that is safe, simple, easy to use, and low in cost. This is an innovative tool for research in natural disasters, geomorphology, and historical disasters.