基于莱特准则的椭圆拟合优化算法

发布时间：2018-04-24 22:30

本文选题：莱特准则 + 椭圆拟合　；参考：《计算机应用》2017年01期

【摘要】：普遍使用的代数距离最小的最小二乘(LS)椭圆拟合算法简单、易实现,但对样本点无选择,导致拟合结果易受误差点影响,拟合不准确。针对此特性,提出了一种基于莱特准则的椭圆拟合优化算法。首先,由代数距离最小的LS法对待拟合曲线进行椭圆拟合;其次,将待拟合曲线上的点与LS法拟合的椭圆的代数距离作为样本点集,在验证该样本点集服从正态分布的情况下,采用莱特准则,将样本点中值大于|3σ|的点判定为野值并剔除,进行多次拟合,直至样本点中无野值;最后,得到椭圆最优拟合结果。仿真实验结果表明,优化算法的拟合误差在1.0%以下,相比同条件下的LS法,其拟合精度至少提高2个百分点。优化算法的仿真结果与其在香烟圆度在线检测中的实际应用验证了此算法的有效性。
[Abstract]:The commonly used least squares ellipse fitting algorithm with the least algebraic distance is simple and easy to implement, but it has no choice of sample points, which leads to the result of fitting being easily affected by error points and inaccurate. Aiming at this characteristic, an ellipse fitting optimization algorithm based on Wright criterion is proposed. Firstly, the least algebraic distance LS method is used to ellipse the fitting curve. Secondly, the algebraic distance between the points on the curve to be fitted and the ellipse fitted by LS method is taken as the sample point set. By using Wright criterion, the point whose median value is greater than 3 蟽 is determined as outliers and eliminated, and then fitted several times until there are no outliers in the sample point. Finally, the elliptical optimal fitting result is obtained. The simulation results show that the fitting error of the optimization algorithm is less than 1.0%, and the fitting accuracy is improved by at least 2 percentage points compared with LS method under the same condition. The simulation results of the optimization algorithm and its practical application in the online detection of cigarette roundness verify the effectiveness of the algorithm.
【作者单位】：上海大学机电工程与自动化学院;
【分类号】：O241.5;TP391.41

【相似文献】