Abstract: A new method for seam tracking using image centroid in gas tungsten arc welding is presented.The centroid of weld pool image from a visual sensor is extracted as the measurement eigenvector.The evolution of the weld position data from the weld pool images can be described through an appropriate state equation and a measurement equation which are based on the image centroid.The advantages of Kalman filtering over least squares approaches is taken to complete the recursive implementation of the image centroid position equation.Also,this Kalman filter (KF) is applied to reduce the error of weld detection caused by the process and the measurement noises,which can estimate the deviation between the arc and the weld accurately.Computer simulations and actual welding experiments have demonstrated that the seam tracking accuracy can be promoted through estimating the weld feature location on the image plane at the next sampling time and adjusting the welding torch position in real time.