In open-pit mines, segmentation of ground point clouds is crucial for the traversability analysis and path planning of autonomous mining trucks. However, mainstream point cloud segmentation methods often suffer from under-segmentation problems due to the heterogeneity of local features in mining environments and the indistinct boundaries between massif and ground. To address the above issues, we propose a novel ground point cloud segmentation method based on quadtree voxel partition. First, the method divides the point cloud of the single frame into voxels using a quadtree structure and adaptively adjusts the voxel composition through point cloud distribution. Subsequently, principal component analysis (PCA) is introduced for iterative parameters optimization and plane estimation within voxels. Finally, a multidimensional function is designed based on the verticality and flatness of the point cloud for likelihood estimation, thereby achieving ground point cloud segmentation. We evaluate the proposed method assessing precision and recall rates under field tests, and the results indicate that the proposed point cloud segmentation method achieves over a 5% improvement in precision compared with mainstream methods.