CSPRS

系統管理

期刊內容

26卷 / 2期 [ 點閱率 : 1828 ]
華藝線上圖書館全期下載(Full Issue Download) : 26(2).pdf
請記下華藝線上圖書館帳號(JPRS000002),密碼(8YqFKj),以便下載文章時使用,感謝您的支持 !

Pages : 57-73    基於資料融合3D LOD2建物模型物件化之研究
論文名稱 基於資料融合3D LOD2建物模型物件化之研究
Title The Establishment of 3D LOD2 Objectivization Building Models based on Data Fusion
作者 邱式鴻、林岑燕
Author Shih-Hong Chio, Tsen-Yann Lin
中文摘要 本研究透過整合2D 地理空間資訊、空載光達點雲、數值高程模型、空載垂直和傾斜航空影像探討大量產製3D LOD2 建物模型物件化之自動化方法。此方法程序分四個階段:(1) 根據每棟建物輪廓內的數值高程模型和光達點雲資料確定shapefile 檔案中每棟建物輪廓相對應的高程和和地面高程,並將其附加到建物輪廓shapefile 檔案中;(2) 將建物輪廓shapefile 檔案轉換為個別由三角形網格化呈現的標準檔案格式物件檔的3D LOD1 建物模型,甚至可以產製具中庭的3D LOD1 建物模型;(3) 從航空垂直和傾斜影像以及全部3D LOD1 建物模型中,確定每一棟3D LOD1 建物模型中每一個三角形面最合適的航空或傾斜影像進行紋理敷貼;(4) 每個完整的3D LOD2 建物模型均通過將其相對應的建物紋理影像進行矩行封包為單張影像並進行資料壓縮完成物件化。本研究中最後將完成之物件化模型導入WebGL 平台,展示進一步之應用。
Abstract This paper discussed an automatic building objectivization method for generating a large number of 3D LOD2 building models by integrating 2D geospatial information, airborne LiDAR point cloud, a DEM, aerial vertical and oblique images. The procedure was divided into the following four stages. First, the elevation and ground floor elevation corresponding to each building in shapefile were determined based on LiDAR point cloud data within each building outline and a DEM, and they were appended into a building outline shapefile. Second, individual 3D LOD1 building models with the standard file format of object file presented by triangulation networks were transformed from every single building outline shapefile into the object file, and even 3D LOD1 building models with any atriums could be produced. Third, from aerial vertical and oblique images as well as whole 3D LOD1 building models, the texture corresponding to each 3D LOD1 building model triangulation was determined by the most appropriate images for texture mapping. Fourth, each complete 3D LOD2 building objectivization model was constructed by splitting and packing its corresponding building texture images into a single image for data compression. Finally, the objectivization models were imported into the WebGL platform for demonstrating the advanced application in this study.
關鍵字 資料融合、三角網、紋理敷貼、矩形包裝、三維建物模型
Keywords Data Fusion, Triangulation Network, Texture Mapping, Rectangle Packing, 3D Building Model
Pages : 75-94    精化多視角影像密匹配及點雲產製
論文名稱 精化多視角影像密匹配及點雲產製
Title Refinement of Multi-view Dense Image Matching and Point Cloud Generation
作者 劉宣萱、趙鍵哲
Author Hsuan-Hsuan Liu, Jen-Jer Jaw
中文摘要 大量多視角影像於密匹配計算處理上較為複雜而繁瑣,且針對每一像對分別計算其初始視差值,耗損的時間成本亦相對增加;再者,具多組像對重疊條件之多視角影像,倘未善加調製其交會幾何,產製之場景點雲即便具有描述幾何多餘觀測特性,然而點位的不精確性以及較大的離散度亦無助於後續空間資訊之應用。對此,本研究提出一系列優化作業模式並區分為三大主軸:建立影像群聚關係、視差傳遞策略和點雲精化策略等。業經兩組實際資料驗證其功效及可行性,說明所研擬方法產製之點雲能有效描述場景幾何,且於兩測試區域之像對計算總量部分,減少約為44%及14%,而針對其時效提升部分則達80%以上。
Abstract Processing large number of multi-view images is complicated and tedious. Also, when matching multiple stereo pairs, it would take long in getting disparity values if each pair is to be processed independently. In addition, redundantly described scene models with low reliability trouble the exploitation of geospatial information. This paper proposes an effective matching strategy featuring in key view selection and clustering, disparity delivery and point cloud refinement to tackle the aforementioned shortcomings. The proposed approach has been tested by two practical data sets and it is proven that both the efficient image manipulation with the computational time reduction more than 80% and quality point cloud generation well depicting the scene geometry highlight the merit of this study.
關鍵字 影像密匹配、多視角、點雲、精化
Keywords Dense Image Matching, Multi-view, Point Cloud, Refinement
12
Page size:
select
Page: of 2
Items 1 to 2 of 4