Due to self occlusion, multiple shots are necessary to reconstruct a 3D model from a range camera. After surfaces from each view are registered into a global coordinate system, they are integrated into a single surface. This process is called 3D integration, which is a data fusion problem. The integration procedure should deal with redundant data, multiple modalities, and be robust to noise.

The goal is to generate surfaces from registered range data by combining several state of the art reconstruction algorithms.

There are various algorithms for multi-view range data integration, including mesh zippering, implicit surface integration and surface reconstruction from an unorganized point. In this work, first two methods were implemented. An implicit surface-based method is more robust to noise and is the basis of most of the surface reconstruction results in this work. Our algorithm is a combination of Curless and Hilton's methods. If the range camera calibration is unknown, then the algorithm is like Hilton's method, where multi-view surface meshes are the integrated. We apply a hole filling process to create a watertight model if the data is not complete. If calibration parameters are available, Curless' space carving method is added so that a watertight model can be generated directly. After surfaces are integrated, a level set mean curvature flow is applied as a final post-processing step. The method can also handle texture integration.

• Y. Sun, J. K. Paik, A. Koschan, and M. A. Abidi, "3D reconstruction of indoor and outdoor scenes using a mobile range scanner," Accepted by IEEE Int'l Conf. on Pattern Recognition, 2002.
  
  

• Y. Sun, D. L. Page, J. K. Paik, A. Koschan, and M. A. Abidi, "Triangle mesh-based surface modeling using adaptive smoothing and implicit texture integration," Accepted by 1st Int'l Symposium on 3D Data Processing Visualization and Transmission 2002.
  
  

• Y. Sun, C. Dumont and M. A. Abidi, "Mesh-based integration of range and color images," in Proc. SPIE AeroSense, The International Society for Optical Engineering, 2000, vol.4051, pp.110-117.