Self-Calibration Methods for Using Historical Aerial Photographs with Photogrammetric Purposes

The historical aerial photographs are taking a fundamental importance for analysis of the territory. Appropriate comparison of photogrammetric surveys of a particular zone carried out in different years allows the identification of geometric changes occurring during the time interval in question. This technique has many fields of application such as evolution of coastlines or landslide monitoring. The main mathematical principles used in photogrammetry are based on the collinearity condition, which allows threedimensional coordinates to be extracted from stereo photographs. Camera parameters and ground control points (GCPs) are required for obtaining a photogrammetric solution. When the parameters that define the camera calibration are not known (this is the most usual thing when we work with historical flight), then they should be evaluated by a process called self-calibrating bundle adjustment. Leica Photogrammetry Suite (LPS) offers several models for self-calibration such as models of Bauer, Jacobsen, Ebner, and the Brown. In this research, a stereo-pair from a photogrammetric flight at a scale 1:5000 was used. Full details of camera calibration data used were supplied. Within the work area, 87 randomly distributed ground points were selected. These points were located on welldefined natural and man-made features in both digital images. Differential global positioning system observations were used to measure ground points. The aim is to study the influence of different self-calibration methods on the accuracy of three dimensional coordinates and Digital Elevation Models (DEMs) generation. We test six photogrammetric projects: (1) without self-calibration, by setting the principal point coordinates to zero, and fixing the value of focal length, which is usually known as a marginal data in the photograph, (2) based on the previous project but adding Bauer’s models as self-calibration, (3) using Jacobsen’s model, (4) Ebner’s, (5) Brown’s, and the best possible project, (6) using the full information about the camera included in the camera calibration certificate. All these projects were computing using 24 and 12 GCPs respectively. Ebner’s orthogonal model, with twelve additional parameters and with 24 GCPs, was the best option for self-calibration, presenting accuracies very close to the projects that used camera calibration data, being their DEMs almost identical.