無人機於數值地形模型建置之路徑規劃方法－以沖積扇為例

Abstract

無人飛行載具(Unmanned Aerial Vehicle, UAV)因具有低成本、高效率之特性，因此越來越多研究將其結合影像建模技術來建置大型地形的數值地形模型(Digital Terrain Model, DTM)。其中沖積扇地形因具有高致災性，因此十分需要長期建置DTM以利逐年的監測。在使用UAV於DTM建置的過程中，UAV的路徑規劃會對整個模型建置的效率有著很大的影響。然而現有的UAV路徑規劃工具，並未針對沖積扇的特性與UAV的限制來規劃出能完整搜集沖積扇影像的路徑。因此，本研究提出一無人機於沖積扇數值地形模型建置之路徑規劃方法，該方法能針對沖積扇的特性與現有UAV的限制規劃出沖積扇影像搜集的飛行路徑。該方法主要包含三大步驟：目標區域切割、起降地點設置、以及近似最佳化路徑計算。在目標區域切割步驟中，會先以能搜集到足夠重疊率的UAV飛行距離為單位，將目標區域切割為數個區塊，藉此確保UAV蒐集到的影像皆具有足夠的重疊率。完成目標區域切割後，UAV起降的地點會需要被手動地輸入，如此才能在計算路徑時將往返起降點的飛行時間考慮進去。最後藉由調整過的蟻群演算法，計算近似最佳化的飛行路徑來充分使用有限的UAV電池。為驗證該方法能有效提升沖積扇的影像搜集效率，本研究將該路徑規劃實際導入沖積扇並與現有之飛行紀錄比較的，結果顯示此方法能減少19% 的影像搜集時間。此外，為驗證該方法確實能協助UAV操作者搜集到擁有足夠覆蓋率之影像，本研究於臺中霧社水庫上游沖積扇進行現地的實驗。實驗結果顯示，該路徑規劃方法確實能協助操作者獲取有足夠覆蓋率之影像，並能建置出完整的DTM。

This research develops an unmanned aerial vehicle (UAV) path-planning method that aims to ensure the required image overlap and optimize the flying routes when applying the UAV for digital terrain model’s (DTM) reconstruction. To collect images on a terrain for image modeling, enough overlap between each collected image must be ensured. In addition, when planning the optimized flying routes for collecting images on a debris fan, the specifications of the debris fan and the limitations of the UAV should both be taken into consideration. Therefore, a UAV path-planning method for image collection is proposed in this research. The developed path-planning method takes a debris fan as an example and refers to the specifications of a debris fan and the limitations of the UAV. The developed method can help the operators to ensure the image overlap through dividing the debris fan into cells by the UAV’s maximum image collection distance interval. The near-optimized UAV flying paths are calculated though applying and modifying a typical path-planning algorithm, the ant colony optimization algorithm. The developed method is validated to be able to help operators to sufficiently use the limited UAV batteries and evaluate the efficiency of the image collection process by comparing the path-planning result with an actual flying log. A site experiment was also conducted for validating the workability of the developed method. The result of the comparison shows that the path-planning method can reduce 19% of the image collection time. It also confirms that applying the method on an actual debris fan can guarantee the required image overlapping and generate a complete DTM without model breaking. In conclusion, this research successfully develops a UAV path-planning method for DTM’s reconstruction.