An omnidirectional imaging system for the reverse engineering of industrial facilities
Digital photogrammetry continues to evolve from specialist applications such as topographic mapping and is rapidly emerging as a highly accessible method for capturing geometric data. A range of general-purpose softcopy photogrammetric systems are now widely available to end users who are thus able to exploit images captured from an increasing number of high-resolution non-metric digital cameras. In parallel with these developments, an increasing diversity of range-imaging systems are being developed to facilitate the rapid acquisition of geometric data. To date, these devices do not offer the resolution, portability or speed afforded by digital cameras. However this thesis anticipates the development of hybrid range and intensity imaging systems. Furthermore through the extension of such systems to facilitate the acquisition of omni-directional imagery the thesis seeks to demonstrate the utility of such data in the rapid documentation of complex objects. In many cases developments in this field have been driven by industrial end-users who have the responsibility to document large and complex structures in order to ensure that they conform to design specification. Such documentation is of critical importance at a wide range of component sizes, from table-top sized structures such as those commonly found in the automotive industries to very large industrial objects such as process-plants, offshore oil platforms or power stations. As digital photogrammetry has matured it has been deployed on projects of ever increasing complexity and system developers have been challenged to produce high precision in ever reducing timescales. This thesis will catalogue the requirements of such end-users active in the field of As-Built Surveys of large industrial structures. In response to these needs the thesis will demonstrate the development and exploitation of omni-directional digital photogrammetry and range imaging systems to enable the creation and exploitation of very large image databases. Furthermore the thesis will demonstrate the extent to which computer vision based analyses of such databases can, in turn, permit precise yet cost-effective documentation of a wide range of industrial facilities.