| 摘要註: |
無人飛行載具(UAV)可於空中進行航照影像的拍攝,於近兩年受到廣大的重視,本研究將利用航照影像與數位影像相關係數法(DIC)來建立三維地形模型。本研究採用空中三角測量原理,其中包含了相機內外方位、控制點之影像座標與空間座標等參數,目前多以人工點選法定位控制點的三維座標,此方式容易造成點選錯誤或是點選位置不精準的問題,故本研究以數位影像相關係數法定位取代人工選點法,研究結果證實DIC 有助於提升控制點定位之精度。本研究也進行了室內實驗,利用相機拍攝平面試體、三維階梯試體與橢圓模型進行控制點數量與誤差探討。根據室內實驗結果得知控制點高程複雜性愈高可以提升相機率定之精度且以7個以上控制點率定相機,其效果比僅用5 個或6 個控制點好,另外,將橢圓試體之高程座標還原後,其高程誤差之標準差為0.1mm,與物距比為2.4/10000。室外實驗為拍攝校園之航照影像來進行相機率定與還原影像三維座標,結果顯示,還原後的高程座標其誤差之標準差約0.136m,與物距比約14/10000,此誤差是比光達雷射(蔡建成,2009)與ImageStation Automatic Elevations 自動匹配(Höhle,2009)來的大。 Unmanned aerial vehicle (UAV) can be used to take aerial photographs, and it hasdrawn widely attraction in recent years. This study will use aerial photographs anddigital image correlation method (DIC) to establish three-dimensional terrain models. Inthis study, instead of manual selection, digital image correlation method is used toobtain the image coordinates of control points. A series of experiments are carried out toconfirm the results that DIC can improve the accuracy of the control point positioning.The experiments are divided into indoor and outdoor experiments. The indoorexperiments used flat specimen, three-dimensional ladder test specimen and oval modelto discuss the influence of the amount of control points on analysis error. According tothe indoor experiments, the high variability of control point elevation can improveaccuracy of camera calibration, and using seven control points to the analysis is betterthan using five or six control points. At last, the result of reconstructing spatialcoordinates of the oval model shows that the standard error of elevation is 0.1mm, andthe ratio of this error to the object distance corresponds to 2.4 / 10000. In outdoorexperiments, the aerial photographs of NUK are taken, and these images are used tocalibrate the camera and reconstruct three-dimensional coordinates. According to theresults, the standard error of elevation coordinate is about 0.136m, and its ratio to the3object distance is about 14/10000. This ratio is larger than that obtained by AirborneLiDAR and ImageStation Automatic Elevations. |