In this paper a new method has been presented to determine the optimum
number of knots for cubic splines. The knot-finding process is based
on the numerical integration of the input curve. The number and the lo
cation of the knots is determined automatically. The method has been a
pplied to a test case and the performance has been compared with two o
ther existing methods. It is shown that fewer knots are retained for t
he portions of the curve having small curvature whereas a larger numbe
r of knots is retained for highly curved portions. The computer time r
equired by our method depends only upon the number of points in the in
put curve and does not depend upon the shape of the curve or the desir
ed accuracy.