A method of path planning based on laser cladding process characteristics was proposed, which can improve the machining efficiency and the quality of surface parts. The effective defocusing amount range of the laser cladding process is determined by the coaxial model of laser and powder interaction and the coaxial powder feeding experiments. The points cloud data of the cladding surface was obtained by the 3D scanner and the point set on the cladding path was obtained by the point clouds slicing method, utilizing equal bow height sparse trajectory point set. And then the NURBS (non-uniform rational B-splines) surface was fitted to the cladding surface and its differential geometrical characteristics were analyzed to determine the vector processing point and biased it to a certain distance to obtain the trajectory point of the gun head. The complex movement of laser gun head with continuous variable altitude is determined by the robot flexible processing characteristics, meanwhile, keeping the laser gun head and the cladding surface processing point normal vector on the same line. From the cladding experiments and the quantitative analysis of the experimental samples, good quality of the cladding layer on the curved surface was achieved. Also, the reliability and practicability of the method were verified, thus providing a good reference for remanufacturing of other surface parts.