At present, the ground of open-pit mines is rugged and bumpy, often accompanied by more dust and soil slopes. In order to obtain more front view of the driveless dumping truck perception system, the installation angle of the lidar needs to be changed during the tooling process. In this paper, an algorithm for automatic calibration and variable scale segmentation of lidar is proposed. First, the driveless dumping truck is parked on a wide ground to ensure that the road ahead and the current position of the driveless dumping truck are at the same level. The current pitch angle and roll angle of the lidar are obtained by analyzing the characteristic normal vector in a square area 10 m in front of the trucl. A symmetrical object is placed directly in front of the lidar, and its centroid deviation is calculated to obtain the yaw angle of the lidar. After the corrected point cloud is obtained, the point cloud data is segmented with variable scale. The point cloud is divided into fine-grained grid and coarse-grained grid, and ground and non-ground point clouds are obtained through granular transformation, plane fitting, and neighborhood smoothing. The scheme proposed in this paper can effectively solve the situation of poor ground environment, strong turbulence, and frequent occurrence of soil slopes and smoke in the mining area, so as to ensure the safe operation of driveless dumping truck in the mining area.