以影像邊緣線資訊優化真實正射影像

Abstract

近年來，隨著無人飛行載具的快速發展，在測量製圖的應用尤為顯著，無人 飛行載具具有高機動性、低成本的優勢，且可以在低空雲下作業，相較於傳統航 遙測載具所受到天候及雲層遮蔽的影響較小。傳統航遙測影像進行正射校正，所 產生之正射影像與地圖具備正射投影、單一比例尺之幾何特性，因此可以在正射 影像上直接獲取幾何資訊，也可以與地理資訊系統應用結合。隨著軟、硬體快速 發展，無論是利用空載航照或是衛星影像根據 DEM 校正產生正射影像都已逐 漸普及。有別於傳統正射影像未對建物之高差位移進行修正，真實正射校正考量 所有地表物體在影像的幾何變形與遮蔽，透過數值地表模型校正後之成果影像可 以展示所有地表覆蓋物的「真實位置」，故稱為真實正射校正。藉由高解析度影 像匹配所獲得的數值地表模型，在邊緣線位置以及高程變化處容易產生扭曲或不 連續的現象，連帶導致其產製之真實正射影像邊緣線扭曲情形。主要原因為點雲 密匹配之誤差，以及點雲內插 DSM 所產生的扭曲的情形，因此需對數值地表模 型進行精化，使遮蔽區的偵測能夠正確且不會有扭曲的情形產生，本研究將利用 原始航拍影像，以及密匹配得到之點雲成果，從原始航拍影像找出真正建物邊緣線位置，並對於 DSM 進行修正。

As the rapid growth of the Unmanned Aerial Vehicle(UAV), the application in surveying and mapping is more and more popular. In contrast with conventional remote sensing and photogrammetric images which are restricted by weather conditions, the advantages of UAV include high mobility, low cost and flying capability beneath cloud. On orthoimages prodecued by photogrammetry, we can acquire lots of spatial information, such as distance, angle and area. Because orthoimages are rectified according to DEM, the relief displacement caused by building can bot be modified. The so called true orthoimages which are rectified by DSM (digital surface model) are free from the distortion caused by the relief displacement. However, DSM generated by stereo dense matching may have mismatches especially along the edge of building roofs, thus reduces the quality of true orthoimages. To overcome this problem, this study proposed a procedure which firstly finds the exact building roof edge from original UAV images, and then the mismatches of DSM are modified to produce high quality true orthoimages.