Curve Fitting Of Random Data Pionts Based On Gived Contour And Application

Curve fitting is an important subject in approxmating theory and geometry modeling. Especially disorder set of points, which is paid more attention to, is becoming hot subject recently. Nowadays, there are at least three kinds of methods: first, least-squares, which has too much calculation,, Second, project these points to a plane, to get a binary image, then transform it.. But this method relies on image resolution ratio. Third, curve equation, using points as restrict conditions. However, this method is afraid of noise.Generally speaking, each method has it's compatible interval, and uncompatible interval. In fact, because this subject is about points out of order, there is no regulation among the points. Searching a common algorithm is a difficult thing. In this paper, based on application, propose curve fitting based on contour. The contour is given by users, who discribe the result forecast depending on their prior experience. It could be a figure, such as a circle; or an image by hand.The algorithm firstly makes edge detection using Sobel operator, to get a set of disorder points, then gives a transformation of these points set to delete noise. Secondly based on given contour, using image matching, it selects special pionts set. At last, does curve fitting using cubic B-spline interpolating curve, and gain object curve. Whether this curve is excellent or not, this paper gives a function to evaluate. If the function value is too high, the curve is considered to be not excellent enough, and should be thrown. Then a new edge detection using Laplacian Operator must be made, till the result is good.The result of experiment shows that, because of the interactivation, the algorithm is more precise than others, especially suit for clinical diagnosis of medical image, which needs precision. And it's run time is shorter, can apply for primary forecasting. On the contrary, duing to the manual participation, this paper's algorithm can gain different output owing to different input. For example, old doctor has more precise experience than young one, so the contour getting from them can gain more precise results. This paper' s algorithm cannot use in some particular system, such as police. At the last of paper, I select some journal recently about my subject to have a comparison. The results say that This paper's algorithm has lower time-complexity, less run-time and more excellent curve.