Fig 1.
(a) Correspondence is established by searching closest points. (b) Registration results of ICP.
Fig 2.
The global reference point and the rotation invariant feature.
Fig 3.
(a) Point sets before registration. (b) The intermediate result of our algorithm when the weight is large. (c) Final registration result of our algorithm when the weight is quite small.
Table 1.
Comparison of 2D simulation results.
Fig 4.
Local amplification registration results for Deer in which rotation angle is 30°.
(a) Registration result of ICP. (b) Local amplification result of ICP. (c) Registration result of our algorithm. (d) Local amplification result of our algorithm.
Fig 5.
Comparison of two registration algorithms for Deer in which rotation angles are 90°, 120°, 150° and 180° (from up to down).
(a) Point sets before registration. (b) Registration results of ICP. (c) The intermediate results of our algorithm when the weight of the rotation invariant is large. (d) Registration results of our algorithm.
Fig 6.
The convergence of ICP and our algorithm for Deer.
(a) Deer with 30° rotation. (b) Deer with 120° rotation.
Fig 7.
Comparison of two registration algorithms for Bat, Hammer and Horseshoe (from up to down).
(a) Point sets before registration. (b) Registration results of ICP. (c) The intermediate results of our algorithm when the weight is large. (d) Registration results of our algorithm.
Table 2.
RMS comparison of 2D point sets.
Fig 8.
The convergence of ICP and our algorithm for Bat.