Multi-view three-dimensional object description with uncertain reasoning and machine learning

Acquiring 3-D object descriptions from a single or multiple images has been a key goal of computer vision. A key issue here is making decision at various stages with diverse and uncertain evidence. Most of the previous work in 3-D object description has been focused on feature grouping, and the decisions were usually made by ad hoc operators. First, two experiments of applying systematic uncertain reasoning and learning for 3-D object description are shown. First experiment with a monocular building detection and description system verifies the idea that the uncertain reasoning and learning will bring better result with smaller efforts for tuning the parameters. In the second experiment, Bayesian inference is applied to a multi-view and multi-modal building description system, where the number of evidence inputs varies according to the number of images used. An expandable Bayesian network (EBN) is proposed for such a situation. In the experimental results, the proposed method shows a superior performance to others. Finally, the Automatic Building Extraction and Reconstruction System (ABERS) is presented. ABERS detects and describes complex buildings which consist of flat or sloping polygonal rooftops. Despite the increased model complexity, the computation is maintained affordable by using multiple images and rough range data. Rooftop boundary hypotheses are generated from 3-D features obtained from multiple images and verified from the range data. Information from diverse sources (multiple images and range data) is combined at various levels with various methods, such as probabilistic height reasoning and hypotheses verification with expandable Bayesian networks. Complex sloping rooftops are generated by finding hips and ridges. Experimental results on complex buildings are shown.