Abstract: Simultaneous Localization and Mapping (SLAM) is a critical technology for the autonomous movement and operation of mobile robots in coal mines. To address the issue of low feature matching accuracy in environments with sparse point clouds, this study establishes a SLAM algorithm based on factor graph optimization and tight coupling of LiDAR and an Inertial Measurement Unit (IMU). A novel feature matching algorithm that integrates scan-to-map (s2m) and feature-to-feature (f2f) matching is proposed. This fusion approach enables adaptation to both feature-rich and feature-weak environments, thereby improving point cloud registration accuracy in challenging conditions with weak features. Simulations were conducted using the public KITTI dataset for validation, and quantitative analysis was performed using the EVO trajectory evaluation tool. Experimental results demonstrate that the fusion of s2m and f2f matching algorithms enhances the global consistency of the generated map. This method effectively mitigates the drift issue associated with s2m matching in feature-weak environments and improves the accuracy and robustness of localization and mapping in underground mine environments with poor features.