B-values processing algorithms is the core of GNSS landing system. In multiple reference consistency check, the rule of maximum likelihood estimation is used in traditional processing method of B-values, so some correlativity exists in the traditional B-values process, which is easy to cause the problems such as the difficulties in distinguishing between fault sources, the low fault detection rate. To solve these problems, this paper studies the B-values processing method based on Kalman filter. And the use of the new processing method can improve the performance of the system through the establishment of filter model. Experiments are done and the results show that the new processing method based on Kalman filter is very effective in fault detection, and the use of the method enhances system availability level. In the end, the fault detection rate is increased by nearly 20%.