Comparison of Redescending and Monotone M Estimator for Surface Roughness Estimation Using Machine Vision

A machine vision system is used to estimate surface roughness of machined surface after grinding, milling and shaping. The estimated roughness is compared with stylus measurements. The estimation is usually done by linear regression which suffers from outliers and hence the estimation is not robust. In this paper we compare redescending and monotone type of robust M estimation for estimation of surface roughness. The results show very good error reduction compared to the methods based on linear regression for redescending type of M estimator.