In applications of the state estimation theory, the state vector usually implies some constraints that can be known in advance. Making full use of these constraints will enable researchers to have a better understanding of the relationship between state elements, and theoretically enhance the accuracy of state estimation.Considering the recent achievements in constrained filtering, a brand new data fusion algorithm was provided for systems with constraints. Using linear equalities as constrained functions, the method was implemented by projecting the Kalman filtering results onto the constrained subspace, and using distributed, optimal weighting fusion to process local filtering consequences. With the assistance of covariance matching technique, sensors with abnormal measurements were eliminated during data fusion. Simulation proves the feasibility and efficiency of the algorithm, which shows better stability than the centralized fusion algorithm.