Abstract: To solve the problem that less useful information is accessible by using magnetron oscillating arc to scan weld seam, the paper proposed a novel deviation prediction method of seam tracking with the utilization of Kalman filter on magnetron rotating arc sensor. The position sensor is used to locate the arc and the signal taken from quarter circle sampling was processed by Kalman filter. The basic principle and the calculation steps of this method were analyzed. The weld deviation information of the space curve was finally obtained. The prediction weighted factor was introduced to predict and to correct the deviation information of the seam. The welding experimental results showed that the proposed method is feasible for automatic seam tracking based on magnetron rotating arc sensor and it also improved the accuracy. The work provides a scientific basis for the theory and algorithm of seam tracking prediction.