In order to improve the accuracy and robustness of seam tracking,a weld seam recognition algorithm was proposed based on constraint Kalman filter.Consecutive weld seam images were acquired with the structure light measure system.Weld seam image corners were measured and located with morphological image processing and center points of laser stripe were extracted by Gaussian fitting method.Taking the position and speed of these corners and center points as the state variables of Kalman filter,the equations of state and the measurement were established.Then according to the position relationship between corners and center points,the state constraint equations were built,and then the results from the Kalman filter were corrected by the constraint equations.The experimental results show that the proposed algorithm could effectively track weld seam and improve the recognition accuracy and robustness.