圆参数估计的相对代数距离法  

Method for least-squares circle fitting based on the relative algebraic distance

作　　者：李雄军[1] 

机构地区：[1]深圳大学理学院,深圳518060

出　　处：《深圳大学学报（理工版）》2006年第2期147-151,共5页Journal of Shenzhen University(Science and Engineering)

基　　金：深圳市科技基金资助项目(199904);深圳大学科研基金资助项目(9820)

摘　　要：构造一个新的残差项,提出一种新的最小二乘圆估计方法———相对代数距离法.仿真实验表明,该法总体性能优于质心法;具有封闭解析解,克服了几何距离法的迭代不收敛问题;抗噪声和粗差点影响性能优于代数距离法,在样本点较少且(或)分布较集中时有较好的拟合精度.The relative algebra distance （RAD） was proposed as a new method for least-square circle fitting by using a new residual error item. Simulation tests show that it has better performance than the centroid estimation method, and the closed-form estimation of the circle parameters without divergence which is the main problem in the geometric distance method. Results also show higher accuracy of parameter estimation under some measurement error and outliers compared with the algebra distance method. This new method shows the best accuracy when the number of measurements is small or at small arcs.

关 键 词：最小二乘拟合 圆拟合 参数估计 误差 

分 类 号：O29[理学—应用数学]