運用區塊建物框細緻化技術精進三維建物模型

Research on Dividing Aggregate 3D Building Model into Individual Ones

為發展三維國家底圖，內政部國土測繪中心已於108年度產製全國LOD1之三維建物模型成果，但其中九成範圍之建物模型成果較粗略且不具分戶特性，不利視覺呈現及後續應用。為拓展後續應用並發展永續更新機制，本研究針對臺灣通用電子地圖之區塊建物框，發展「以地籍資料分戶技術為主、高差分割技術為輔」的高效率建物框細緻化方法，並透過人工檢視編修方式，進一步提升成果之品質及成功率。本研究選擇臺中市4個不同建物分布類型的行政區域作為測試區。實測發現，建物密集之城區及城郊混合區，初始分戶成功率可達70～85%；而建物分布稀疏之郊區，初始分戶成功率僅有約40%。輔以人工檢視編修後，可將郊區之分戶成功率大幅提升至75%。後續將擴大試辦範圍並確立作業流程，以低成本高效率方式逐步更新全國LOD1三維建物模型成果。

In order to upgrade Taiwan National Base Map from 2D to 3D, nationwide LOD1 3D building model is completed by the National Land Surveying and Mapping Center in 2019. However, nearly 90% 3D building models are aggregate block models of adjacent buildings, which means the models may not meet the expeatation in terms of visual impact and application. For the enhancing of the application potential and the development of the sustainable update mechanism, the high efficiency dividing method against the aggregate building block of Taiwan e-Map is developed through this project. Based on cadastral data classification technology and supplemented by height difference segmentation technology, the aggregate building blocks are divided through automated method first, and are processed through manual editing for further improvement. This project choice 4 Townships as test area and empirically evaluates the success rate and data quality. In suburban and urban mixed areas, success rate of blocks refinement with cadastral maps is about 70%～80%. In suburban area, success rate is about 40% initial and upgrade 75% by introducing manual editing. The project using empirical research has confirmed the flow of data processing to be workable and a cost-reduction way.