%0 Journal Article
%T 圆柱透视投影轮廓的隐式方程描述和拟合方法<br>Implicit equation description and fitting method for cylinder perspective contours
%A 谢颖
%A 杨向东
%A 芮晓飞
%A 任书楠
%A 陈恳
%J 清华大学学报（自然科学版）
%D 2016
%R 10.16511/j.cnki.qhdxxb.2016.22.025
%X 为了提高从视觉图像中识别并提取圆柱透视投影轮廓的鲁棒性和精度, 将由直线和椭圆组成的投影轮廓曲线用含约束的二元高次隐式方程进行描述。采用该描述方式可将轮廓曲线拟合问题通过线性近似转化为含有非线性约束的非线性优化问题。针对该问题求解时, 先粗略地估计出曲线参数的初值, 再利用初值筛选出内点, 最后采用罚函数法将问题转化为求解无约束问题从而求得该优化问题的解。由于该描述方式引入了轮廓中直线和椭圆的相互约束, 采用该描述方式的轮廓曲线拟合方法比采用独立的直线或椭圆识别方法提取轮廓, 具有更强的抗噪能力。仿真和实物图像的实验结果表明, 该方法能有效提高圆柱投影轮廓曲线拟合的鲁棒性和精度。<br>Abstract：A constrained implicit equation is developed to enhance the robustness and precision of perspective contour extraction for cylinders. The contour curve fitting problem is converted to a nonlinear optimization problem containing nonlinear constraints by a linear approximation. The problem is then solved with a rough estimate of the curve parameters and the noisy points filtered to find reasonable points based on the rough estimate of the curve parameters. This constrained problem is then converted to an unconstrained problem using the penalty function method to find the solution. This method takes advantage of the constraint between the lines and the ellipse to give better anti-noise performance for perspective cylinder contour extraction than separate methods using lines or ellipses. Tests of synthetic and real images show that this method effectively improves perspective contour curve fitting for cylinders.
%K 图像识别
%K 轮廓曲线
%K 隐式方程
%K 曲线拟合
%K 非线性优化
%K &lt
%K br&gt
%K image recognition
%K contour curve
%K implicit equation
%K curve fitting
%K nonlinear optimization
%U http://jst.tsinghuajournals.com/CN/Y2016/V56/I6/640