The International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences. Vol. XXXVII. Part Bl. Beijing 2008
973
The 3 parameters of the 3d planes corresponding to the
segments, i.e. the two angles that define the direction of the
normal vector of the plane, and the distance between the plane
and the origin, are computed on-the-fly as the averages of
values computed in the participating pixels. They are expressed
in the same polar coordinate system that the range image is
based on.
3. SENSOR LOCALIZATION USING EXTRACTED
PLANES
The localization of the laser scanner is based on measuring
distances to a minimum of three non-parallel planar surfaces
(that intersect in non-parallel lines) in an already mapped
indoor scene. The first scan maps the scene with respect to the
scanner coordinate system; i.e., the parameters of the extracted
planes are computed in the scanner coordinate system at its
initial position. In the next scan, new ranges to these planes
with known parameters are measured. The position of the
scanner with respect to the previous coordinate system is then
determined by intersecting the measured ranges.
The determination of the scanner position with respect to the
previous position requires a minimum of three corresponding
planes in the two successive scans. Generally a larger number
of planes are extracted by the segmentation algorithm, and a
search procedure for finding corresponding planes is needed.
The following sections describe the plane intersection and the
search procedure for corresponding planes.
(3)
The solution of the set of equations given in (3) provides the
coordinates of S 2 in the coordinate system of S,. By repeating
this procedure at each consecutive scan, the position of the
scanner with respect to the coordinate system of the scanner at
the previous position can be computed.
To transform the scanner positions to a single coordinate
system, e.g. the coordinate system of the first scan, rotations
between the coordinate systems of the scanners at successive
scans should also be taken into account. If the positions of the
scanners are required in a reference coordinate system, then the
position and orientation of the scanner at the first position with
respect to the reference coordinate system should be known.
Assume that a third scan is performed, and the position of the
scanner at S 3 is to be determined in a reference coordinate
system O; we have:
12 13 12 2 12 D T} _,0
P P ~ ^ -s 3 — n •R.jjRqiSj
22 23 22 2 12 R R 0
P P ~ n s 3 — n k 12 k 01 s 3
32 33 _ „32 _2 _ „12 TJ r» „0
P P — n • s 3 — n • R 12 R 01 s 3
(4)
1.1 Scanner position determination
The position of a point can be computed by measuring its
distance to a minimum of three non-parallel planes with known
parameters. Fig. 4 shows three non-parallel planar surfaces
scanned from two positions S t and S 2 . As can be seen, if planes
PI, P2, and P3 are shifted by -p", -p 12 , and -p 13 respectively,
they intersect at point S 2 . The equation of a plane i in the
coordinate system j of the scanner at position s 7 is written as:
p ,J =nfx iJ +4y iJ +n«z ,J 0)
where Ry denotes a 3D rotation from the coordinate system i to
the coordinate system j. The rotation matrices can be derived
from the parameters of corresponding planes in every pair of
successive scans, except for Roi , which is the orientation of
the scanner at its initial position with respect to the reference
coordinate system, and has to be measured in advance (if the
navigation is to be computed in a reference coordinate system).
In practice, the scanner can be levelled by using the bubble
level and levelling screws. In such case, the z axis of the
scanner will always remain in the upward direction.
Consequently, the 3D rotation matrices will be simplified to
rotations around z axis only, which can be easily derived from
the differences in the parameter 0 of the corresponding planes.
where p ij is the distance of the plane i from the origin of the
coordinate system j, ,n‘{, n'j are the coordinates of the
normal vector n of the plane i in the coordinate system j, and
x‘ J f y'j 5 z‘ J are the coordinates of a point x on the plane i in the
coordinate system j. The short form of the equation can be
written as:
where . denotes the dot product of the two vectors. To compute
the position of the scanner at position S 2 we take the
parameters of three non-parallel planes in the coordinate
system of Si, shift them according to their distances to S 2 , and
find the intersection. This can be expressed as:
Fig. 4: Determining the position of S 2 by measuring distances
to three planes.