Abstract: The problem of automatically tracking the sync-phase axis on the seismic cross-section images via pattern recognition and AI approaches is one of the major topics in seismic data explanation. This paper presents a methol which employs AR model to represent the sync-phase axis and track the axis automatically by a special data structure and searching technique. Compared with the results described in [1] and [2], this method is capable of repairing the broken points, thus keeping the continuity of the axis. It can also provide useful information about the ends of the axis as well as the broken strata and their locations. The structure of the proposed system can be provided with an interface for knowledge input, thus making it possible to track intelligently.