Recently, the LIGO-Virgo collaboration reported their first detection of gravitational wave (GW) signals from a low mass compact binary merger GW170817, which is most likely due to a double neutron star (NS) merger. With the GW signals only, the chirp mass of the binary is precisely constrained to 1.188 +0.004 −0.002 M , but the mass ratio is loosely constrained in the range 0.4 − 1, so that a very rough estimation of the individual NS masses (1.36 M < M 1 < 2.26 M and 0.86 M < M 2 < 1.36 M ) was obtained. Here we propose that if one can constrain the dynamical ejecta mass through performing kilonova modeling of the optical/IR data, by utilizing an empirical relation between the dynamical ejecta mass and the mass ratio of NS binaries, one may place a more stringent constraint on the mass ratio of the system. For instance, considering that the red "kilonova" component is powered by the dynamical ejecta, we reach a tight constraint on the mass ratio in the range of 0.46 − 0.59. Alternatively, if the blue "kilonova" component is powered by the dynamical ejecta, the mass ratio would be constrained in the range of 0.53 − 0.67. Overall, such a multi-messenger approach could narrow down the mass ratio of GW170817 system to the range of 0.46 − 0.67, which gives a more precise estimation of the individual NS mass than pure GW signal analysis, i.e. 1.61 M < M 1 < 2.11 M and 0.90 M < M 2 < 1.16 M .