True orthoimage generation has become one of the most investigated research topics motivated by the growing
technology of high resolution image acquisition. Conventional orthophotos are based on differential rectification in their
production. Unfortunately, in large scale urban imagery differential rectification produces a serious problem in the form of
double mapped areas called “ghosting effects”. True orthoimage generation techniques try to remove this ghost effect by
detecting the occluded or obscured areas, marking them as blank, filling them from the neighbouring images and finally treat
shadowed areas. This paper presents a method for true-orthoimage generation from high resolution aerial imagery. The
method compromises three main steps: (i) image orientation based on collinearity equations/bundle block adjustment, (ii)
Digital surface model (DSM) using semi-global matching (SGM) technique, and (iii) true-orthoimage generation. The
obtained true-orthoimage is a rigorous one with no self-occlusions, ghost effects and multiple texture mapping. The use of
semi-global matching for DSM generation has developed promising results for orthoimage generation. Using an accurate
DSM generated from image itself refined from occlusions and outliers eliminates the serious ghost effect with no need for
subsequent steps for occlusion detection and elimination. |
