圆曲线拟合算法中的粗差探测与识别

针对传统的圆曲线拟合算法抵御粗差影响能力较差的现状,该文提出了一种圆曲线拟合数据的粗差探测方案。该方案基于Pratt代数拟合算法,首先构造衡量拟合数据优劣的质量指标,据此迭代剔除数据中的异常值,然后在既得圆心估值的基础上,重新评估已被标记的可疑观测值,识别出真正的粗差并将其剔除。实验结果表明,该方法能够准确识别出数据中的粗差,有效提高圆曲线拟合的精度。

Detection and identification of outliers for circle fitting

For the traditional geometric and algebraic circle fitting algorithms cannot resist the influence of outliers,a method of detecting the outliers based on Pratt algorithm from data points for circle fitting was proposed in this paper.Firstly,the quality indicator for every point was constructed by which the suspicious observations were detected iteratively.And then based on the reliable estimates,all the original observations were re-checked,and the real outliers were identified and removed.Computation results show that not only the outliers can be detected and identified correctly by using the proposed method,but also the accuracy of circle fitting can be improved efficientlv.