| Automatic surface reconstruction has been an important research topic in digital photogrammetry for many years. Its central part is to solve the matching problem automatically. Most algorithms used to solve the matching problem can be categorized as either area-base techniques or feature-based techniques. Over the past years, feature-based matching has gained more and more popularity for it is the method that supports activities in object recognition and image understanding. Hence it fits well with the new definition of modern digital photogrammetry.; Recent research in digital photogrammetry and computer vision demonstrate the success of using linear features for solving the matching problem, which results in data sets of 3D linear features in the object space. While more research in linear feature matching is expected, there are only a few investigations regarding the construction of 3D linear features for surface reconstruction. There is clearly a lack of research in reconstructing surfaces by using 3D linear features in the object space.; This research is concerned with the reconstruction of surfaces from 3D linear features, in particular 3D lines, in the object space. Intermediate data processing will include 3D line segmentation, 3D line grouping, 3D line linking, classification into ground and non-ground lines, plane finding and plane fitting. Planes are checked whether they are real physical planes. If this is the case the boundary of the plane are determined.; of plane formats, is proposed. The advantages over traditional surface representation, DEM and TIN, are discussed. It is expected that the concept of surface construction using 3D lines in this research may be as well applicable to object recognition, in terms of moving from a classical search in the 2D image space to a new 3D domain in the object space.; The recommended surface using 3D lines will facilitate subsequent vision tasks such as object recognition. Spatial reasoning in object space is much more powerful than reasoning in 2D image space. |
