無人機影像建模模擬日照陰影之研究

Abstract

本研究以無人機作為地面資訊收集工具，透過自行航拍興趣區域之影像，提出一套可行的資料處理及轉換程序，以產出該區任意時段之陰影模擬成果。從UAV影像處理到產出陰影模擬成果包括以下5項主要流程：1.影像匹配、2.點雲過濾分類處理、3.點雲建模、4.模型格式轉換、5.陰影範圍推算等流程。在影像匹配階段，以航空攝影測量原理為基礎，對不同電腦視覺軟體進行驗證，比較其點雲分類成果在陰影模擬上之優缺點後，建議以Agisoft Metashape軟體進行影像匹配。由於影像匹配產生之點雲仍帶有部分雜訊需剔除，因此以影像內真實陰影資訊作為參考資料，推估地物高度，手動進行點雲過濾。過濾後屬於建物的點雲可產製三角形網格模型(Mesh Model)，但要能進行陰影分析還必須轉換為面格式模型(Multi-patch Model)，才能以ESRI ArcGIS Pro軟體進行陰影模擬，產出陰影分佈區域。陰影模擬流程之實驗以國立臺灣師範大學和平校區校園為實驗區，以無人機航拍高重疊影像，並透過VBS-RTK GNSS測量控制點及檢核點三維座標，以驗證本文提出之陰影模擬流程。成果展示以兩種呈現方式模擬真實環境下對於掌握陰影範圍之需求，分別為：1. 產出校園單棟建物在一日之內對周遭地面造成的陰影範圍與累計遮蔽時長，以此模擬在真實環境下單一建物在長時段下對周遭地區產生之遮蔽時間與範圍，進一步作為後續日照權等社會環境影響議題之參考資料；2. 推估校園內一塊地面廣場一日之中受陰影覆蓋的起迄時間，展示對於特定感興趣之空間，探討此空間任意時段下受陽光陰影覆蓋之時間區間，以規劃該區之時間與空間應用。

A practical procedures to simulate shadow coverage at the interested area at specific time is proposed in this research, based on the DEM and 3D building models generated by aerial photographs taken by an unmanned aerial vehicle (UAV). The proposed procedures includes 5 major steps: (1)image matching, (2)point cloud filtering and classification, (3)triangulation for mesh model, (4)3D building models transformation, (5)shadow coverage simulation. At the first step, two computer-vision-based software, Agisoft Metashape and Pix4D Mapper, are compared based on the computed interior and exterior orientation. The results shows that Agisoft Metashape has better performance, and therefore, it is chosen for image matching and point cloud generation. Since noise points are inevitable, it is necessary to filter point cloud before triangulation for mesh models. The real shadow appears on the specific image is manually measured as the reference to estimate the height of the building. Those points are higher than the building’s height are considered as noise will be filtered out. After filtering, the points classified as buildings are triangulated as mesh models. These mesh models have to be transformed to multi-patch model and imported into ArcGIS Pro for generating shadow coverage.The HePing campus of the National Taiwan Normal University is chosen as the study area in this research. The high-percentage-endlap aerial images are taken by a UAV. The 3D coordinates of ground control points and check points are measured by a VBS-RTK GNSS receiver. The Agisoft Metashape is chosen for image matching, point cloud classification, and mesh model generation. The shadow coverage is simulated for two scenarios: (1)For a specific building, simulating it’s shadow coverage during the whole day to analyze the impact to it’s surroundings. (2)For a ground area, calculating the time under the shadow to analyze it’s application.