When autonomous vehicles are driving in complex environments, the presence of uneven ground and obstacles can impact the safety of the vehicle. Traditional obstacle detection algorithms overlook the impact of sloped terrain on vehicle path planning. In environmental perception, determining whether a slope is passable has become a hot and challenging issue. This paper proposes a 3D LiDAR-based ground segmentation algorithm. This method relies on data obtained from experimental equipment for terrain information collection systems ahead of the vehicle, enabling it to distinguish between ground points and non-ground points in steep terrain. Experiments show that this method achieves good results in both static and dynamic slope detection.