"Proceduralization" of scanned buildings

Internal

• Laser scans (LIDAR)
  • Street view
  • Aerial view
• Images

• RANSAC

• http://www.c3technologies.com/

• LAS Tools: LAStools: converting, viewing, and compressing LIDAR data in LAS format. Contains sample LAS files (mostly terrains).

• Leibniz Universität Hannover
• Aimatshape

• MeshLab: MeshLab is an open source, portable, and extensible system for the processing and editing of unstructured 3D triangular meshes. The system is aimed to help the processing of the typical not-so-small unstructured models arising in 3D scanning, providing a set of tools for editing, cleaning, healing, inspecting, rendering and converting this kind of meshes.

• Arc3D
  • Manual
  • Images can be uploaded and then depth maps are created (dense), as well as a point cloud.
  • Depth maps can be integrated with the MeshLab software suite (freely available).

• Unstructured textured polygons used in this paper by Johannes Kopf et al., from MSFT research.

• Textured triangles in VRML format.
• Sample data made available by UNC contains three zip files of about 98, 124, and 168 MB. Each file is a different scene and contains camera calibration and poses, video frames, and 3D models produced using our real-time system (see Pollefeys IJCV 2008). The 3D models are in VRML format (just vertices, face indices, and texture coordinates). There is also an example Matlab script that shows how to manipulate the data.
• The models have been saved to: \\matrix\Scratch\urban\proceduralization\data\UNC

• These screenshots have been generated from the models using Cortona3D viewer.

• Stabilizing 3D modeling with geometric constraints propagation. Michela Farenzena and Andrea Fusiello.
• PDF
• Approach:
  • Input: Manually created triangulation that represents an object composed of planar faces.
  • Automatically create geometric constraints (coplanarity, parallelism, orthogonality, angles equality) and form a Geometric Constraint System.
  • Solve GCS to obtain final model.

• A Connection between Partial Symmetry and Inverse Procedural Modeling. Martin Bokeloh, Michael Wand, Hans-Peter Seidel.
• PDF
• Docking sites are an interesting concept. Can they be extended to non-perfect information? Can docking sites be identified in noisy point sets and used as a reference to build the grammar and/or reconstruction?

• Non-local Scan Consolidation for 3D Urban Scene. Qian Zheng, Andrei Sharf, Guowei Wan, Yangyan Li, Niloy J. Mitra, Baoquan Chen, Daniel Cohen-Or.
• PDF

• SmartBoxes for Interactive Urban Reconstruction. Liangliang Nan, Andrei Sharf, Hao Zhang, Daniel Cohen-Or, Baoquan Chen.
• PDF
• Summary:
  • Takes advantage of orthogonality and regularity.

• CityFit: High-quality urban reconstructions by fitting shape grammars to images and derived textured point clouds. Bernhard Hohmann, Ulrich Krispel, Sven Havemann and Dieter Fellner.
• PDF
• Summary: The goal of the CityFit project is to reconstruct the facades of 80% of the buildings in the city of Graz fully automatically. The challenge is to establish a complete workflow, ranging from acquisition of images and LIDAR data over 2D/3D feature detection and recognition to the generation of lean polygonal facade models.
  • The input data for CityFit consist of highly redundant road side photographs and LIDAR scans acquired by Microsoft Photogrammetry. The LIDAR 3D point cloud is not sufficient for direct facade reconstruction. It allows, however, to derive the main orientation and the rough structure of the facade very reliably. For examining the preprocessing results a point cloud viewer was developed (section 3). The LIDAR data is used together with information extracted from the road side photographs.
• Question for us: Do we want to use information from the images to compute grammar?

• Detailed Real-Time Urban 3D Reconstruction From Video. Pollefeys et al.
• PDF
• Input: Video + GPS/INS
• Process:
  • Camera pose estimation from GPS + INS
  • Sparse scene analysis: determine 3 orthogonal sweeping planes
  • Depth estimation: compute depth maps
  • Depth map fusion: reject erroneous depth estimates and remove redundancies from data
  • Create triangular mesh for each fused depth map (using a multi-resolution, quad-tree triangulation algorithm (top-down approach))

• Take input data from different sources (e.g., Arc3D, Pollefeys IJCV'08, etc).
• Use non-local consolidation (Zheng et al., 2010) to improve and complete point set.
• Use docking sites (from Bokeloh et al., 2010) to partition point set?