Presented in this thesis is a method for constructing "conic spline curves" from a sequence of 2D point data with measuring error. A conic spline is composed of line, circle and general conical curves (parabola, ellipse, hyperbola). The proposed conic spline fitting method has two phases:
(1) breaking phase : data points are partitioned into several groups so that points in a group can be fitted by single(line or circle) segment.
(2) blending phase : two adjacent segments are smoothly joined by defining a conic segment at the join. Conic curve segments are used as blending curves because they provide smoothness together with simple mathematical properties.
In the breaking phase, the data point are first approximated by a polygon, and then they are partitioned into groups by applying simple heuristic rules.
An algorithm is proposed to find a conic segment that is tangent to the base segments while closely approximating the data points.