A Method to Deal with Recognition Deviation Based on Trajectory Estimation in Real-Time Seam Tracking

In real time seam tracking process, recognition deviation is likely to occur due to various kinds of noise (e.g., reflection and scattering ... ). In this paper, a method to correct deviation via replacing the deviation points with estimated points is proposed. Firstly, according to the characteristic of the real-time seam tracking process, standards for recognition deviation are developed to classify deviation points, and an abnormality judgment strategy is discussed. Then, a method of trajectory estimation is given and the detected trajectory of an abnormal deviation will be replaced with the estimated trajectory. Finally, experiments are conducted to prove the performance of the proposed method.