This paper proposes a laser SLAM algorithm based on double-layer global pose map optimization to address the low accuracy and efficiency of traditional laser SLAM loop detection and the long optimization time of large-scale loop back-ends. This algorithm improves the LINK3D descriptor and calculates landmarks accordingly, achieving pre optimization of global keyframes for hypersubgraphs through joint constraints of landmarks and pose. Secondly, the pose factor graph is further optimized according to the pre optimization results to reduce the number of iterations and improve the efficiency and accuracy of the back-end optimization. Experiments on multiple sequences in the KITTI dataset show that the SLAM algorithm proposed in this paper reduces the backend pose optimization time by an average of 17.2% compared to traditional graph optimization methods. For large-scale scenes with many loops, the SLAM algorithm reduces the average translation error by 13.2% and the average rotation error by 13.3%, thereby verifying its effectiveness.