|Ph.D Student||Shragai Ziv|
|Subject||Aerial Triangulation Algorithms in Modern Photogrammetric|
|Department||Department of Civil and Environmental Engineering||Supervisor||Professor Emeritus Yerach Doytsher|
|Full Thesis text - in Hebrew|
This research focuses on one of the most critical stages of the photogrammetric processing workflow, which is the Aerial Triangulation phase. The role of Aerial Triangulation is the recovery of the position and orientation of the camera, at the precise time of capturing of each image, while extracting the locations in space of all of the tie points.
The commonly used answer for the Aerial Triangulation task is the "Bundle Adjustment", which constitutes a statistically optimal Solution. This Solution has already been in use for decades in the photogrammetric community, and has also been recently recognized in the Computer Vision community as a leading technique that usually is applied as a final step for every reconstruction process. The Bundle Adjustment is based on an iterative refinement of all parameters, including the location and orientation of each image captured, the locations of all tie-points, as well as constraints of various additional external observations and measurements such as control points.
Despite its many benefits, Bundle Adjustment has one major drawback, which is being highly sensitive to the existence of fairly accurate initial values for all of the parameters. These should be practically close to the global solution for the process to converge.
The main objective of this research is dealing with the above issue. The research dealt with a wide range of cases expected to arise when dealing with modern aerial mapping projects. Starting with a case where both the location and the orientation are known with limited accuracy, through a situation in which only the location data are known, to a situation where both the location and the orientation parameters are unknown. The dissertation describes three distinctive approaches for solving aerial triangulation problems, under different constraints and varied level of complexity.
The research results, as reflected from the experiments described in this work, show that the integration of developed algorithms in the photogrammetric processing workflow may lead to a more efficient utilization of resources required to perform a classical or modern photogrammetric mission. This newfound efficiency can result in the lowering of the entrance threshold to the aerial mapping field, and in a more simplified photogrammetric process.