Aiming at the problems of low welding efficiency， poor accuracy and unstable operation of welding robots in high-end manufacturing field， a method of kinematics parameter identification of welding robots based on improved Fourier series trajectory was proposed. Firstly， a kinematic model of the welding robot was established based on the MDH （modified denavit-hartenberg） criterion， and an error model of the welding robot was established based on differential kinematics. Secondly， the optimal improved Fourier series trajectory was found through pattern search algorithm， and 50 pose points with the smallest number of conditions were selected as the experimental pose set. Finally， a RLS-DEH （recursive least squares and differential evolution hybrid） algorithm was proposed， which can effectively improve the accuracy of parameter identification. The above method was validated using KUKA’s welding robot as the experimental object. The experimental results showed that after， substituting the optimal pose set obtained using the improved Fourier series trajectory method into the error model and identifying it using the RLS-DEH algorithm， the average absolute position error of the welding robot was reduced from 1 mm to within 0.05 mm， which improved the accuracy by 95% compared with the previous calibration， proving the efficiency and practicality of the proposed method.