Abstract: Objectives To address the problem of visual beacon visibility degradation and pose estimation accuracy decline caused by attitude deviation during autonomous underwater vehicle (AUV) approach to recovery cage, a novel binocular visual guidance method integrating optimized light array layout and pose estimation is proposed. Methods First, a "3+3" dual-layer non-coplanar light array layout is designed, which enhances the diversity of spatial geometric constraints by deploying light sources on different depth planes. Second, the binocular visual pose estimation algorithm is improved by introducing a nonlinear optimization step based on binocular reprojection error on the basis of traditional triangulation combined with SVD initial value solution. Results Simulations and ground experiments demonstrate that within a range of 0.5~10 m and yaw angles of ±60°, the translation error of the method is controlled within 4% of the relative distance, and the rotation error is less than 3.2°. Even when only 3 light sources are visible, stable solution can be maintained, and the pose solvability probability is significantly superior to single-layer layouts. Conclusions The proposed method significantly improves the robustness and accuracy of AUV autonomous recovery terminal guidance under complex viewing angles through the combination of spatial depth layout and optimization algorithm.