During the flight test and route operation of large civil aircraft, it is necessary to inspect the appearance of the aircraft for damage around the aircraft. At present, the aircraft around inspection mainly adopts manual winding method, which has high cost and low efficiency, and is prone to human factors such as missed detection and misdetection. Therefore, the research on intelligent surface inspection method is an urgent task. Compared with other industrial detection tasks, this project currently has no public data set, and the real appearance surface image is blurred and diversified. In this paper, an aircraft surface damage image data set is preliminarily established, and an aircraft surface inspection engineering method based on YOLO V3 is proposed. The framework can solve the common deep learning network intelligent detection or recognition of known types of damage, has a strong tolerance to the unknown damage, making the algorithm has greater flexibility and adaptability. Firstly, the YOLO detection network was used to roughly obtain the damage location and damage classification of aircraft appearance. Secondly, according to the characteristics of different damage types, the appropriate traditional algorithm was used to obtain the more accurate damage location in the image block. Finally, the quantitative analysis was carried out according to the refined results. A method is proposed to solve the problem of intelligent detection of known category damage by machine deep learning network, which has strong tolerance and flexibility and adaptability to unknown damage. The experiment and practice show that the algorithm proposed in this paper can overcome the disadvantages of the traditional manual visual inspection, reduce the maintenance and maintenance cost of aircraft during flight test and operation, and is of great significance to the maintainability design of aircraft.