Abstract: Feature matching-based Simultaneous Localization and Mapping (SLAM) algorithms rely on sufficient geometric feature information. In challenging scenarios such as underground coal mines and long corridors, geometric features may be insufficient, which leads to mismatches in SLAM systems. To address this problem, a laser SLAM algorithm for underground coal mines based on fusing point cloud intensity constraints was proposed. On the basis of traditional geometric feature extraction, the reflectance intensity information of point clouds was fully utilized, and texture features were additionally extracted as constraints for pose estimation, thereby improving the accuracy and stability of pose estimation in scenarios with weak geometric information. A loop closure detection algorithm based on the Intensity Scan Context (ISC) descriptor was designed, which used intensity distribution features to achieve more robust scene matching, thereby improving the global consistency of pose graph optimization. The algorithm was validated through tests in narrow corridors and real underground coal mine environments. The results showed that, compared with mainstream laser SLAM algorithms, the proposed algorithm exhibited significant robustness in feature-degraded scenarios. The maximum horizontal localization error was 0.426 m and the maximum elevation error was 1.801 m. The overall localization accuracy and system robustness were significantly improved.