Kalman filters used in rocket ballistics' real-time computation are mainly based on observing model with position and velocity, which requires to convert the ranging and bearings data to position first in view of external ballistic metrical data. Therefore such algorithms can't take full advantage of range, bearings, and velocity measurement data in exterior trajectory measurement. In view of this, an adaptive Kalman filtering algorithm was proposed, which was based on the Taylor expressions and established observing model in station horizontal coordinate system. Then some real metrical data was used to test and verify it. The results indicate that this algorithm can filter noise effectively.