Abstract: To address the challenges of interference susceptibility during assembly hole extraction and low efficiency of pose estimation in cabin docking, an efficient hole pose determination method is proposed. Firstly, the YOLO algorithm is employed for the rapid detection and localization of the cabin end-face. Simultaneously, secondary and tertiary arc segment matching, based on projective invariance, is applied to enhance matching efficiency and rapidly acquire precise two-dimensional (2D) coordinates of the assembly holes. Subsequently, a two-stage matching strategy integrating epipolar and geometric constraints is utilized to achieve the high-speed matching of assembly hole images captured by a binocular camera. Building upon this, a binocular spatial cone model is established for pose estimation. A multi-hole pose optimization model is then constructed, employing a genetic algorithm (GA) to calculate the optimal attitude vector of the cabin. Experimental results demonstrate that the proposed two-stage matching method improves matching efficiency by 76.5%. The relative errors for translation and attitude angle measurements are constrained to within 2.71% and 2.00%, respectively, while the overall pose estimation speed is accelerated by 31.44%. These findings thoroughly validate the effectiveness and efficiency of the proposed method.