Abstract: The mixed estimation is necessary for fusion of various heteroscedastic multi-source observation models in geodesy. Because the additional constraints and the sample information play an unequal role in the estimation process, we need to establish a new weighted adjustment criterion to balance the influence of prior constraints and observation information on parameter estimation. Firstly we regard multi?source observation data as observation information and some random constraints information, use ellipsoid approxima-tion to describe bounded uncertain information, and establish an adjustment criterion based on the minimum trace of outer ellipsoid characteristic matrix. Then, we propose a new method of observational information fusion and a method of calculating optimal weights, which makes the weighted mixed estimation method effective in geodetic data processing. Finally, the validity of the algorithm is verified by an example, and the relationship between the set membership estimation solution and the weighted mixed estimation is illustrated.