Classification of River Habitat Using Large-Scale Particle Image Velocimetry with Unmanned Aerial Vehicle

Generally, physical habitat, i.e. flow velocity, flow depth, dimension of sandbar, was surveyed by using handhold equipment. The survey processes were time and cost consuming. This study attempted to apply an unmanned aerial vehicle (UAV) with a camera to record the images of a river. The images were used to measure the sizes of coarse particles and the dimensions of sandbars. Moreover, the surface velocities of river were obtained by large-scale particle image velocimetry (LSPIV) from the images. The information can be used to classify the river habitat. It is expected to simplify the complex process of traditional habitat survey. A field experiment was performed to quantify the habitat survey. The sizes of coarse particles and dimensions of sandbars were measured by a tape measure and images analysis from the UAV. Two tracing particles including the natural and artificial particles were used in the LSPIV analysis. The flow velocities measured by LSPIV were compared with the velocities measured by the handhold Acoustic Doppler Velocimeter (ADV). The results showed the error of particle sizes was about 5.4%. The deviation of velocities measured by LSPIV with artificial particles and ADV was about 0.08m/s, and the deviation of velocities measured by LSPIV with natural particles and handhold ADV was about 0.06m/s. The results of the field experiment showed the LSPIV analysis provided similar flow information with the ADV. The UAV classified the river habitat in a large field. It provided a more efficient and accurate method to obtain river habitat information than using the conventional equipment.一般物理性棲地調查多以手持式量測儀器獲取單點流速水深及底床粒徑等棲地資訊，重覆施測直到將整個分析區域量測完畢，往往需耗費許多人力及時間。本研究嘗試以無人飛行載具進行現地拍攝透過航拍影像分析河道棲地中粗顆粒之粒徑，並結合大尺度質點影像量測法取得河流表面流速，進行河道棲地型態分類，及相關棲地指標計算，希望能取代傳統單點量測方式，減少現場繁複作業，提供棲地調查更有效率之方法。 實驗時於河道中實際量測若干粗顆粒之粒徑尺寸，以便與航拍影像計算值做比較。流速分析則分別利用人工施撒Ecofoam及水流表面自然產生之泡沫做為大尺度質點影像量測法之追蹤質點，並與手持式ADV現場實測流速值進行比較。實驗結果顯示利用影像計算河道中粗顆粒之粒徑與實際量測值誤差為5.4%，而人工施撒粒子進行LSPIV分析及自然追蹤質點進行Unseeded LSPIV分析與ADV實測流速之平均差異分別為0.08m/s與0.06m/s。現場試驗證實，以LSPIV分析與手持式ADV相較，可得到相似之流場資訊，UAV之分析結果可將大範圍河道進行棲地型態分類，相較於以往傳統儀器之棲地調查法，UAV為更有效率及精確之棲地調查方法。摘要 i Abstract ii 目錄 iii 表目錄 v 圖目錄 vi 第一章 前言 1 1.1 研究動機與目的 1 1.2 本文架構 2 第二章 文獻回顧 4 2.1 無人飛行載具 4 2.1.1 無人飛行載具種類 5 2.1.2 無人飛行載具多元應用及精度分析 6 2.2 大尺度質點影像量測法（Large-Scale Particle Image Velocimetry，LSPIV） 8 2.2.1 LSPIV於不同載具之應用 9 2.2.2 不同追蹤質點之選擇 13 2.3 棲地型態分類 15 第三章 研究材料與方法 21 3.1. 基本原理介紹 21 3.1.1 數值地表模型（Digital Surface Model，DSM） 21 3.1.2 質點影像量測法PIV 25 3.1.3 棲地型態分類法 28 3.2. 實驗設備 33 3.2.1 無人飛行載具 33 3.2.2 Pix4Dmapper軟體 36 3.2.3 手持式ADV（Acoustic Doppler Velocimeter） 38 3.2.4 追蹤粒子 40 3.3. 實驗說明 42 3.3.1 空間資訊之座標誤差比較 43 3.3.2 LSPIV分析流場 49 3.3.3 棲地型態分類 55 第四章 結果分析與討論 56 4.1. 航拍影像之誤差討論 56 4.1.1 空間資訊之座標誤差比較 56 4.1.2 粗顆粒表面粒徑及沙洲尺寸量測誤差比較 60 4.2. LSPIV現地實驗結果分析 66 4.2.1 第Ⅰ區（LSPIV與手持式ADV） 68 4.2.2 第Ⅱ區（LSPIV與Unseeded LSPIV） 77 4.2.3 第Ⅲ區（Unseeded LSPIV與手持式ADV） 83 4.3. 棲地分析 87 4.3.1 棲地型態分類 87 4.3.2 M1、M2、M3指標計算 91 第五章 結論與建議 98 5.1. 結論 98 5.2. 建議 99 參考文獻 100 附錄一 10