Aimed at the statistical property estimation of measurement noise in the adaptive Kalman filtering, a method based on α-β-γ filter and second order mutual difference estimation is proposed to estimate measurement noise variance.By constructing a virtual redundant measurement sequence with α-β-γ filter, the second order mutual difference method only adaptive to redundant measurement is extended to single measurement.Based on which, the calculation precision is improved through selection of data and optimization of α-β-γ filter.Experimental result indicates that: compared with traditional adaptive Kalman filter, this method can compute the measurement noise variance more accurately, and restrain the filtering divergence.