The composite wing skin has the characteristics of large size, complex shape and easy rebound. It can’t be used for nondestructive testing by traditional methods such as machine tools. But the robot has the characteristics of flexibility and intelligence, which provides a new idea for nondestructive testing. A scanning path optimization method of large wing skin continuous surface is proposed to solve this kind of problems. Dual robots equipped with ultrasonic scanning equipment, adopting the strategy of two-time inspection: the composite surface is reconstructed by the first scanning, so the accuracy of the second transmission nondestructive inspection is improved. According to the shape of the wing, a general scanning strategy parallel to the stringer is proposed. The points are grouped by the least squares method according to the curvature, the path is optimized by the hybrid genetic LM algorithm. The algorithm means the improved genetic algorithm is used for heuristic global optimization and the LM algorithm is used for deterministic local optimization, so that the optimal scanning path can be obtained efficiently. Then, the simulation is carried out in RoboDK, and the robot is equipped with  ultrasonic detection end to scan the skin. Finally, the precision of the optimized path is verified by the laser scanner on the robot. Simulation and experiment results show that, compared with traditional detection methods, the average detection efficiency of this method is improved by 9.2%. It meets the constraints of ultrasonic detection.