For describing urban scenes and modeling building objects, strategies are presented considering segmentation, classification, and geometric modeling of point clouds. Hierarchical voxel structures are introduced to exploit neighborhood relation in unstructured point clouds. A new local shape descriptor, advanced feature extraction methods, and segmentation based on Gestalt principles are proposed. Local and global graphical models are employed for optimizing the segmentation and classification. Results from three datasets are compared with those of state-of-the-art methods.
«
For describing urban scenes and modeling building objects, strategies are presented considering segmentation, classification, and geometric modeling of point clouds. Hierarchical voxel structures are introduced to exploit neighborhood relation in unstructured point clouds. A new local shape descriptor, advanced feature extraction methods, and segmentation based on Gestalt principles are proposed. Local and global graphical models are employed for optimizing the segmentation and classification. R...
»