Summary: J.E. Banta, L. M. Wong, C. Dumont and M. A. Abidi, "A Next-Best-View System for Autonomous 3D Object Reconstruction," IEEE
Transactions on Systems, Man and Cybernetics, Vol. 3, No. 5, pp. 589-598, September 2000.
590 IEEE TRANSACTIONS ON SYSTEMS, MAN, AND CYBERNETICS--PART A: SYSTEMS AND HUMANS, VOL. 30, NO. 5, SEPTEMBER 2000
Fig. 2. (a) Range image, and (b) model reconstructed from the surface data
contained in (a) and the occluded data inferred from (a).
in a 3-D bitmap (occupancy grid) can take on a value of 0 (unoccupied)
or 1 (occupied). (More information on 3-D modeling can be found in
C. Range Data Registration
To register range data, we must first transform the range data from
2 1/2-dimensional to 3-D by calculating the transformation from one
coordinate space to another. We can easily shift a greyscale range
image into three dimensions by separating the grey levels into binary
bit planes and occupying the voxel corresponding to each pixel and
appropriate bit plane. We then apply a transformation that maps a
ranged point to its corresponding Cartesian coordinates.
To restrict view positions to points outside of the model and to en-
sure model visibility, the sensor is always placed on the view sphere
and oriented toward the approximate center of the object. Because the