Improving the robustness of circular curve fitting with equality constraints using the median absolute deviation method
Chuan Hu, Daqin Ren, Huarong Li, Shuai Zhang, Guobing Pan · Survey Review · 2025
Circular curve fitting with equality constraints using robust nonlinear weighted total least squares with equality constraints (RNWTLSEC) based on the iteratively reweighted method is affected by the initial values. To improve its robustness, the RNWTLSEC-MAD (median absolute deviation) algorithm is proposed. The algorithm first applies MAD to residuals from nonlinear weighted total least squares with equality constraints (NWTLSEC) to identify and eliminate outlier-containing observations. Then, the NWTLSEC estimates from the remaining observations are used as the initial values for RNWTLSEC estimator. The results of two experiments show that RNWTLSEC-MAD detects more outliers than the other four methods.