Abstract: In order to overcome the time-consuming and laborious characteristics of traditional land leveling measurement methods, LiDAR technology is proposed to reconstruct farmland terrain. The hardware platform of the system is built by HDL-32E LiDAR, and the data acquisition program is compiled by C++ language. On this basis, the data collected by the LiDAR is calibrated, and the transformation methods of different coordinate systems in the farmland terrain reconstruction system are studied. Meanwhile, based on the minimum denoising method, a denoising method with mean difference is designed, which is more suitable for the denoising of farmland terrain point clouds. This is the case. The system accuracy is evaluated by comparing the number of units measured by RTK and LiDAR in the region with large fluctuation of farmland and without slope. Finally, the interface display, application and accuracy evaluation of the vehicle-mounted farmland terrain reconstruction system are realized. The results show that the farmland topography obtained by LiDAR is more abundant and accurate in the range of of 10°~15°、25°~30°. The projection area approximation of the reconstructed farmland topographic model point cloud data and the original farmland topographic point cloud data can reach 93%. It shows that the research method in this paper can be applied to the reconstruction of farmland topographic environment. At the same time, it provides a theoretical reference and basis for the future work of land fine leveling.

Key words: LiDAR, farmland terrain, land leveling, information fusion, reconstruction