In off-road environments, detecting hazardous terrain such as cliffs, steep slopes, and negative obstacles is the key to safe driving for autonomous vehicles. In this study, we propose a hazardous terrain detection method named HTD. This innovative method entails the sequential fusion of multiple frames of LiDAR point clouds and applying Bayesian generalized kernel (BGK) to generate a dense digital elevation map (DEM). Subsequently, the detection of hazardous terrain is performed through the analysis of features derived from the DEM and the accumulated point cloud. In addition, we release a LiDAR hazardous terrain detection dataset, which includes typical hazardous terrain data (cliffs, steep slopes, negative obstacles). The dataset will be made publicly available at https://github.com/guanglei96/HTD.