An accurate and fast image-based visual servo error compensation method for 3d concrete printing robots in complex scenarios

3D concrete printing (3DCP) has been widely adopted in construction for its high design freedom and excellent manufacturability of complex structures. However, traditional methods rely on preprogrammed paths or rigid template matching, making them vulnerable to dynamic errors from material rheology, environmental disturbances, and motion step loss. These errors severely degrade interlayer printing accuracy, final product quality, and forming consistency. To address this, we propose a dynamic error compensation framework based on image visual servoing, integrating high-precision semantic segmentation and dynamic trajectory optimization. We design an attention-based feature fusion module to focus on key contour regions, an improved refined segmentation head for precise mask extraction, and a structure-aware loss function with dynamic path weighting. A contour center trajectory extraction algorithm maps corrected visual trajectories to robot joint space for high-precision control. Experimental results show that the proposed method achieves a mean Intersection over Union of 97.39% in normal environments and 97.02% in complex environments for mask segmentation. On complex lattice windows, cumulative errors are reduced by 85.67% compared to traditional methods, with dimensional accuracy controlled within 0.676%. For conventional components under normal lighting, the mean absolute errors in X and Y directions are only 1.1743 mm and 0.8825 mm, respectively. These findings demonstrate the robustness and effectiveness of our approach, providing technical support for high-precision 3DCP.