Recognition and 3D-reconstruction of objects from images using a priori information
Alexei Zakharov, Arkady L. Zhiznyakov · 2014
The probabilistic approach of three-dimensional reconstruction of the visual environment of urban scenes from satellite and aerial images is presented. The mathematical model of the reconstructed objects is presented. Contour images of models of reconstructed objects are shown. Conditional probability of occurrence of recognizable signs and reconstructed objects are used in the model. Hough transform is used for feature extraction. The approach aim is to find maximum a posteriori probability of the synthesized model. A maximum a posteriori probability is using Monte Carlo Markov chain scheme. Possible transitions between the models for the iterative search are presented in the paper.