無地面控制點之近景攝影測量幾何改正與誤差分析

Abstract

在災害監測或調查的應用中，例如地震後的建物倒塌、山坡地監測或土石崩塌調查等，若能快速取得災害現場的三維空間資訊，將可以作為後續進行災害搜救、災因調查，或是災後復原重建等之參考。傳統的災害調查大都仰賴專業人員親赴現場，以簡單的測量工具搭配紙筆進行災害現場資訊之記錄，工作不僅繁重且易出錯，相對而言，近景攝影測量(close range photogrammetry)以不同拍攝角度拍攝災害現場之影像，除了可忠實記錄災害現場資訊之外，亦可減輕外業工作之負擔。一般近景攝影測量的施測方式需事先於觀測目標之間或其外圍佈設網形好的控制點，經由相機獲取影像後量取控制點影像坐標，利用多重影像幾何交會透視原理，解算透視中心外方位參數，進一步再利用前方交會獲取物體的三維位置、形狀、大小或建立模型。然而災害現場通常不易事先建立控制點，或是事先建立的控制點已經遭受破壞，因此如何在沒有地面控制點資訊的情況下進行近景攝影測量為本研究之主要探討目的。此外，近幾年來直接定位(direct geo-referencing)製圖系統的發展，亦可以不需現場控制資訊，直接利用系統本身的GPS定位及IMU方位資訊，獲取待測物之空間三維坐標。本研究將針對近景攝影測量中各種可能的控制來源，分析其對影像幾何改正(方位解算)及物空間解算精度之影響，並提出對災害勘查人員有幫助的建議。實驗結果顯示，在無任何控制點下的內業求解方式是可行的，而無控制點的成果通常都會好於有控制點的成果，若無控制點的場景(例如災區)有約制條件的提供，則僅需2個精度高的約制條件就可以發揮其約制功效。

Close range photogrammetry(CRP), has become an accepted, powerful and readily available technique for engineers and scientists who wish to utilise images to make accurate 3-D measurements of complex objects. Traditional CRP requires establishment of good ground controls to guarantee high precision. Photos with control points are taken, and exterior orientation parameters(EOPs) of camera exposures are solved by perspective geometry. Then, the position, shape and size of the object in three dimentional space were got by means of intersection.Compared with other surveying tools, traditional surveying need operators carry heavy instruments, but CRP can reduce the inconvenience caused by this. Besides, in disaster management, it is usually hard for people or vehicles to reach the disaster scenes, or we cannot find corresponding satellite images. Thus, we will use non-control point CRP as a basis. In this paper, we illustrate the procedureal strategy under no ground control point information and compare the process with the normal CRP case. We consider three forms of control information: ground control points, control cameras, and geometric elements. By means of analysing simulated data and real data, we have tested the feasibility of our proposed method. The result shows the EOP precision is not consistent with the existence of control points. But with the control points information, we may get more accurate 3D position with both the same reliable precision. As for the control constraints, they indeed can improve the EOP precision under no control point CRP case, but they increase the precision of object point coordinates with limits.