The fitting of curves in computer aided geometric design is generally regarded as an optimisation problem. According to the application, conditions to be satisfied can make the problem difficult to solve with classical methods. Stochastic ones, such as gen
QFigure 1 : Gcomposite typical curve (3 segments)2
3.Fitting using a genetic algorithm
3.1. Statement of the problem
The composite curve we have defined allows the definition of different curve profiles or curve templates, throughthe combination of typical curves with a monotone increasing and/or decreasing curvature. This possibility comesat first sight with the user’s choice according to the considered data points set. The case that we chose to treat ina first step consists in fitting a curve on a data points set (Pi , i = 0..N) whose underlying shape presents aincreasing curvature variation.
We also considered the number m of curve segments as fixed. It can be indeed empirically chosen by the useraccording to the shape complexity. In the same manner, the point Q0 location and the total curve arc length S (S = Sk for k = 1..m) are fixed in advance as :
Q0 = P0 and S = Pj-1 Pj for j = 1..N
Provided that the data points do not present significant outlier points and their density is relatively high, the soestimated curve arc length gives, as proven by our results, good curve solutions.
3.2. Coding and fitness function
The genetic code we have defined consists in the following parameters set :
(W0, 0), ( 1, 1), ( 2, 2), ……..( m , m)where the first parameters subset corresponds to the tangent direction and curvature at Q0, and the following
subsets determine the m curve segments by the relations :
Sk = ( k - k-1) S , with k 0, 1 for k = 1..m-1 and 0 = 0 , m = 1
Wk = k k-1 Sk
where k-1 is the curvature at the ( k-1)th curve node, and where k is a parameter related to the curvature gradientof the kth segment. The previous variable aims to set the Wk values properly by taking into account the level ofcurvature at each curve node and the different segment arc length, so that the curvature variation is globallygrowing. Another advantage is to provide the search domain for each segment parameters subset with a relativehomogeneity.
Fitness of generated curves is determined according to the following function of the quadratic deviations:
Fitness 2
i 0N2di Pi C(ui) i 0N
where C(ui) is the curve point, normally defined as the nearest of the data point Pi. Calculating this point cannot be operated analytically and then must be numerically processed, which canincrease the global calculation cost, considering the number of data points and the number of iterations processedin a genetic algorithm. So, we developed an analytical curve deviation approximation method, based on an arclength parametrization. At first, we search the data points nearest to the curve nodes and define in that way, thecurve segment in the neighborhood of the considered data point. The arc length separating the correspondingcurve point from the beginning segment endpoint (Fig.2) is then written as :
k Si lk
LSk
where lk is the cumulated chord length, calculated on the considered data points segment, up to Pi , and thei
Lk the total length of this segment.

th curve kth data polygonal

segment
Abstract Adaptive T-spline Surface Fitting to Z... Computational Geometry and Object Modeling—curve,...styling, NC machining, GIS, and computer vision....
Abstract Conceptual design in the automotive ...for a design application where creativity should ...The resulting 2D curve is projected back into 3D...
deforming 3d shapes in biomedical and styling ...Abstract In contrast to machined mechanical parts,... Computational Geometry and Object Modeling—Curve,...
deforming 3d shapes in biomedical and styling ...Abstract In contrast to machined mechanical parts,... Computational Geometry and Object Modeling—Curve,...
热门文档