טכניון מכון טכנולוגי לישראל
הטכניון מכון טכנולוגי לישראל - בית הספר ללימודי מוסמכים  
M.Sc Thesis
M.Sc StudentLevin Shahaf
SubjectClose-Range Photogrammetry Using DTM as Control Surfaces
DepartmentDepartment of Civil and Environmental Engineering
Supervisor Professor Sagi Filin
Full Thesis textFull thesis text - English Version


Abstract

The application of terrestrial photogrammetry for documentation, measurement and reconstruction of sites is widely known; this is due to the ease of data collection, availability of imaging devices, the ability to avoid physical contact with mapped objects, and the existence of well-established photogrammetric documentation methodologies. Nonetheless, there are cases where site complexity requires the application of alternative photogrammetric-driven approaches. Traditionally, photogrammetry is based on use of ground control points. This well-established methodology is limited however to areas where such control networks exist or can be constructed with a sufficient level of accuracy. Alternative, line-based geo-referencing models suffer from the same fundamental problem of relying on singular, identifiable control entities.


The present study shows that by utilizing the entire object-space, in its "surficial" form, more information can be harnessed for geo-referencing. It thus presents a surface-based photogrammetric model and its consequent application for site documentation of scenes acquired from close range. The proposed model benefits from the use of the entire object-space, and the existence of well-proven surface matching methodologies. While surface-based geo-referencing of aerial images have been presented in recent years, the differences from terrestrial models, both in application and implementation, make it almost impossible to apply to the terrestrial case. A proper solution requires studying this subject from a terrestrial modeling perspective. Three major issues that differentiate terrestrial models from aerial ones are identified. These include overcoming the inherent occlusions, managing the different scales and resolution, and finding first approximations.


Aiming for a general model that can be applied for site documentation, this research explores the implications of those special circumstances on the registration. The model, application is demonstrated on both high and low-resolution surface information in order to investigate the effect of surface details on the quality of pose parameter estimation. Results show that accurate pose parameters can be obtained without the need for designated well-defined control targets. The model is than applied to various sites, including an archeological site and a hillside scene. Difficult accessibility to both sites made the establishment of steady control network impossible, thus making them perfect test subjects. The results show that the level of accuracy in the image alignment of the test sites was comparable to a point-based solution.