In this paper， a new stochastic model updating method is proposed， which combines the random hybrid perturbationGalerkin method with the improved cross-model cross-mode technique. This method effectively alleviates the impaction of limited measurement data and uncertain measurement errors on model updating. Considering the uncertainty of the measured modal data， a new stochastic updating equation with update coefficient vector is established based on the improved cross-model cross-mode method. Using the hybrid perturbation-Galerkin method to solve the stochastic updated equation， the update coefficient vector is obtained. The statistical characteristics of the update coefficients can then be determined. The numerical results of the simply supported beam show that the proposed method can effectively deal with the relatively large uncertainty in the actual measurement data，and shows relatively strong stability in the case of different modal combinations， and has a higher computational efficiency than the Monte Carlo method. Considering the rank deficit， the improved cross-model cross-mode method proposed in this paper can get better updating results than the cross-model cross-mode method. The experimental results of the frame show that the new method can simultaneously modify the stiffness and the quality of the structure， and the updated model can be used to obtain modal data consistent with the measured results， thus verifying the effectiveness of the proposed method.