<p>This paper proposes a novel self-positioning cutting control strategy for tunnel boring machines (TBMs), integrating modern computer vision and deep learning technologies to achieve precise cutting control of the TBM under different working conditions. By using monocular vision technology combined with deep learning algorithms, a method for detecting the six-degree-of-freedom pose of the TBM was developed, which can monitor and calculate the pose changes of the TBM in complex environments in real time. To achieve automatic adjustment and precise control during the cutting process, a two-layer fuzzy adaptive backstepping control method with specified performance is proposed. This control method combines the nonlinear processing ability of fuzzy control with the precise adjustment characteristics of adaptive back-stepping control, enabling efficient real-time adjustment of the TBM’s actions in complex and dynamic working environments, ensuring the precision and stability of the cutting process. In addition, the study applied this control strategy to the cutting head trajectory tracking control experiment under different body poses of the TBM. In the experiments, the TBM performed cutting head trajectory tracking control under two different body poses, verifying the applicability and reliability of the method. The experimental results show that, under these two body poses, the maximum contour errors of the cutting head were 42 mm and 45 mm, with errors of approximately 3.51 % and 3.79 %, respectively. These results demonstrate that the proposed self-positioning cutting control method for the TBM can effectively control cutting accuracy in practical applications and has high engineering practical value.</p>