Modern aerial refueling aircraft and its aerial refueling technology have become an important measure to enhance the mobility and strike capability of aviation forces, and are highly valued by countries all over the world. With the continuous development of computer image and machine learning, it is possible to locate and identify the refueling cone sleeve in real time in the process of aerial refueling docking. At present, the mainstream method is target recognition based on deep learning, but the premise of deep learning is to cut a large number of data samples. Aiming at the problem that deep learning requires a lot of manpower and energy to cut training samples, this paper proposes an automatic cutting method of cone sleeve training samples based on the combination of color features and circular features to realize the automatic pre-processing of massive sample data Cutting, greatly improve work efficiency, lay the foundation for UAV autonomous refueling. © 2022 IEEE.