In mine operations, the safe operation of transportation equipment is crucial to ensure the safety of miners and the efficiency of mine production. However, it is notable that there is little research on perception technology for unstructured environments such as underground mining tunnels. The underground mining environment is characterized by its intricate nature, with narrow passageways, dim lighting, and complex spatial topological structures. Large-scale mining trucks operating in such environments have a restricted field of view and pose a serious safety hazard. In this paper, we propose an underground mining obstacle detection method based on 3D light detection and ranging (LiDAR) technology to augment the environmental perception capabilities of mining vehicles. This method uses point cloud data collected by LiDAR as input, employing an improved random sample consensus (RANSAC) to segment rugged ground points. Additionally, an innovative point cloud processing module for tunnel walls and the application of Euclidean clustering and obstacle recognition strategies ensure accurate obstacle detection. Experimental results demonstrate that the proposed method achieves a detection accuracy of over 95% within a 50-meter region of interest, and the running time is kept within 0.14 seconds on an ordinary computer. The effectiveness of the proposed method is discussed across varying distances, numbers, and tunnel types, revealing satisfactory outcomes and robust applicability. The proposed efficient method meets the requirements of underground mining truck obstacle detection, making a substantial contribution to underground unmanned production.