This thesis presents a generative statistical approach for the fully automatic three-dimensional (3D) extraction and reconstruction of unfoliaged deciduous trees from wide-baseline image sequences. Tree models improve the realism of 3D Geoinformation systems (GIS) by adding a natural touch. Unfoliaged trees are, however, difficult to reconstruct from images due to partially weak contrast, background clutter, occlusions, and particularly the possibly varying order of branches in images from different viewpoints. The proposed approach combines generative modeling by L-systems and statistical maximum a posteriori (MAP) estimation for the extraction of the 3D branching structure of trees. Background estimation is conducted by means of mathematical (gray scale) morphology as basis for generative
modeling. A Gaussian likelihood function based on intensity differences is employed to evaluate the hypotheses. A mechanism has been devised to control the sampling sequence of multiple parameters in the Markov Chain considering their characteristics and the performance in the previous step. A tree is classified into three typical branching types after the extraction of the first level of branches and more specific Production Rules of L-systems are used accordingly. Generic prior distributions for parameters are refined based on already extracted branches in a Bayesian framework and integrated into the MAP estimation. By these means most of the branching structure besides tiny twigs can be reconstructed. Results are presented in the form of VRML (Virtual Reality Modeling Language) models demonstrating the potential of the approach as well as its current shortcomings.
«This thesis presents a generative statistical approach for the fully automatic three-dimensional (3D) extraction and reconstruction of unfoliaged deciduous trees from wide-baseline image sequences. Tree models improve the realism of 3D Geoinformation systems (GIS) by adding a natural touch. Unfoliaged trees are, however, difficult to reconstruct from images due to partially weak contrast, background clutter, occlusions, and particularly the possibly varying order of branches in images from diffe...
»