空照彩色立體像對中人工建築物萃取之研究

Building Extraction from Aerial Color Stereopairs

徐偉城(Wei-Chen Hsu)

《航測及遙測學刊》 5卷3期 (2000/09) Pp. 15-42

https://doi.org/10.6574/JPRS.2000.5(3).2

摘要

三維地理資料的重要性日益增加，其對如三維都市模型、景觀模擬、無線電基地台之選址等均極重要。以目前台灣的地理條件和科技發展，航空攝影測量仍為最重要的製圖方法。但隨著科技發達，其做法也從類比航測進步到解析航測，目前更朝向自動化的數值航測進行。目前雖然已能自動化的處理空中三角測量，但在地物的自動測繪方面，仍然不能得到令人滿意的成果。因此人工建築物的自動化重建，是目前數值攝影測量的重要研究課題。由於人工建築物是直立於地表上的地物，具有相當的特性，因此要自動化的重建，就必須盡量的萃取出其特徵資訊，包括點、線、面及其高程。本研究使用空照彩色立體像對，針對像空間中房屋特徵進行萃取。主要利用影像灰度值和幾何條件的性質：先進行色彩模型轉換，分離不同之光譜反應值，並以區塊成長法萃取均調之屋頂面，再結合邊緣線偵測的結果，以本研究所提出之「邊緣線強度極大值法」找出屋角點之位置，並決定房屋之外型。最後使用本研究設計之「CLR特徵匹配法」，求得房屋邊界之視差，並檢核萃取之合理性。

關鍵字

數值攝影測量 ； 特徵萃取 ； 影像匹配

並列摘要

Three dimensional geographic data is getting important. For those applications such as 3-D city model, landscape visualization, selection of base station for telecommunication, the topographic data is indispensable. Photogrammetric mapping technology is still the major approach for 3-D topographic mapping. The technology has evolved from analogue to analytical one and now toward automation. In the automation, aerial triangulation has been successful and in operation. On the other hand, the automated determination for ground object is still in the early stage. Reconstruction of man-made building, in particular, is the major task in the photogrammetric automation. On the contrary to natural objects, man-made buildings are with regular boundaries and elevation drops. Thus, it is needed to extract corner points, edges, regions, and elevation distribution in reconstruction for man-made buildings. This investigation uses aerial color stereoscopic images to reconstruct those buildings. Considering spectral and geometric characteristics, the components of the reconstruction are: (1)color transformation, (2)region growing, (3)edge detection, and (4)corner detection. Finally, we develop a ”CLR matching” algorithm to determine and to validate the disparities for the building boundaries.

並列關鍵字

Digital photogrammetry ； Feature extraction ； Image matching