Photo-Guided Modeling

The photo-guided modeling phase uses our drawing and modeling tools, along with the native SketchUp tools, to build the 3D model from photographs. This process begins by drawing 2D sketches of footprints or rooftops directly on the photos. This establishes the base of the 3D volumes, which will be created by extruding these footprints. The images projected from the known camera poses serve as guiding visual cues during sketching, extrusion, and tapering.

After the cameras and the floor are aligned to the ground plane, the user can select images from the input set and look at the 3D scene through their respective camera frustums. The user then sketches on the ground plane. That sketch is properly projected into all of the remaining views. The user can select our 2D drawing tools and outline the visible footprint of the building. Our tools include a rectangle, polyline, circle/ellipse, or spline, all constrained to a 2D plane (e.g., floor plane or any polygonal face). Our system allows the user to switch from one viewpoint to another during sketching to add points from corners that are occluded in the current view. The footprints are shown in black in the figure below. Since the cameras are calibrated, a footprint drawn in one viewpoint will appear registered in the other viewpoints and will be aligned with the sparse 3D point cloud.


The drawn footprints are then extruded using the standard SketchUp push-pull interface. This naturally forms 3D volumes that are embedded into the scene.


The images associated with the frustums are then projected to create phototextured 3D models, as shown here.