A method to register the position and posture of a rigid object which
moves in three-dimensional space, without points and/or vector matchin
g between before and after the movement is described. In the method, a
group of coordinates of representative points of the object in the in
itial state, P-i(x(i), y(i), z(i)) (i = 1 to n), and those after the m
ovement P-j(x(j), y(j), z(i)) (j = 1 to n), are used for the input dat
a without knowing their point correspondences. Firstly, the object is
translated from the initial position so that its centroid coincides wi
th one of the goal positions. The difference in posture of the object
is corrected by rotating round the centroid by the torque which attrac
ts it to the goal posture. It is demonstrated that repulsive forces to
each point of the object from all points at the goal posture, whose m
agnitude is the square of the distance between the points, satisfactor
ily produce such torque. The practical procedures for determining the
position and posture using this principle are described, together with
their characteristics and examples of applications to actual data.