next up previous
Next: References Up: A Volumetric Method Previous: 6 Results

7 Discussion and future work

We have described a new algorithm for volumetric integration of range images, leading to a surface reconstruction without holes. The algorithm has a number of desirable properties, including the representation of directional sensor uncertainty, incremental and order independent updating, robustness in the presence of sensor errors, and the ability to fill gaps in the reconstruction by carving space. Our use of a run-length encoded representation of the voxel grid and synchronized processing of voxel and resampled range image scanlines make the algorithm efficient. This in turn allows us to acquire and integrate a large number of range images. In particular, we demonstrate the ability to integrate up to 70 scans into a high resolution voxel grid to generate million polygon models in a few hours. These models are free of holes, making them suitable for surface fitting, rapid prototyping, and rendering.

There are a number of limitations that prevent us from generating models from an arbitrary object. Some of these limitations arise from the algorithm while others arise from the limitations of the scanning technology. Among the algorithmic limitations, our method has difficulty bridging sharp corners if no scan spans both surfaces meeting at the corner. This is less of a problem when applying our hole-filling algorithm, but we are also exploring methods that will work without hole filling. Thin surfaces are also problematic. As described in section 3, the influences of observed surfaces extend behind their estimated positions for each range image and can interfere with distance functions originating from scans of the opposite side of a thin surface. In this respect, the apexes of sharp corners also behave like thin surfaces. While we have limited this influence as much as possible, it still places a lower limit on the thickness of surface that we can reliably reconstruct without causing artifacts such as thickening of surfaces or rounding of sharp corners. We are currently working to lift this restriction by considering the estimated normals of surfaces.

Other limitations arise from the scanning technologies themselves. Optical methods such as the one we use in this paper can only provide data for external surfaces; internal cavities are not seen. Further, very complicated objects may require an enormous amount of scanning to cover the surface. Optical triangulation scanning has the additional problem that both the laser and the sensor must observe each point on the surface, further restricting the class of objects that can be scanned completely. The reflectance properties of objects are also a factor. Optical methods generally operate by casting light onto an object, but shiny surfaces can deflect this illumination, dark objects can absorb it, and bright surfaces can lead to interreflections. To minimize these effects, we often paint our objects with a flat, gray paint.

Straightforward extensions to our algorithm include improving the execution time of the space carving portion of the algorithm and demonstrating parallelization of the whole algorithm. In addition, more aggressive space carving may be possible by making inferences about sensor lines of sight that return no range data. In the future, we hope to apply our methods to other scanning technologies and to large scale objects such as terrain and architectural scenes.

Acknowledgments

We would like to thank Phil Lacroute for his many helpful suggestions in designing the volumetric algorithms. Afra Zomorodian wrote the scripting interface for scanning automation. Homan Igehy wrote the fast scan conversion code, which we used for range image resampling. Thanks to Bill Lorensen for his marching cubes tables and mesh decimation software, and for getting the 3D hardcopy made. Matt Pharr did the accessibility shading used to render the color Buddha, and Pat Hanrahan and Julie Dorsey made helpful suggestions for RenderMan tricks and lighting models. Thanks also to David Addleman and George Dabrowski of Cyberware for their help and for the use of their scanner. This work was supported by the National Science Foundation under contract CCR-9157767 and Interval Research Corporation.


next up previous
Next: References Up: A Volumetric Method Previous: 6 Results