In the osmotic dehydration process of food, on-line estimation of concentrations of two components in ternary solution with NaCI and sucrose was performed based on multi-functional sensing technique.Moving Least Squares were adopted in approximation procedure to estimate the viscosity of such interested ternary solu-tion with the given data set.As a result, in one mode of using total experimental data as calibration data andvalidation data, the relative deviations of estimated viscosities are less than ～ 1.24%.In the other mode, by taking total experimental data except the ones for estimation as calibration data, the relative deviations are less than±3.47%.In the same way, the density of ternary solution can be also estimated with deviations less than ± 0.11% and ± 0.30% respectively in these two models.The satisfactory and accurate results show the ex-traordinary efficiency of Moving Least Squares behaved in signal approximation for multi-functional sensors.