The land precision leveling technology could improve farmland micro-topography and distribution uniformity of irrigation water and soil nutrients, save water and fertilizer and increase both production and income. To improve the working efficiency of GNSS-controlled land leveling system, a method of farmland regional planning based on spatial clustering was explored. Firstly, a topography contour map was generated by using terrain rapid measurement method. Farmland altitude data was obtained by the topography map. Secondly, the altitude data was divided into different categories by using K-means clustering algorithm, and the data in the space was marked by each category. Thirdly, the discrete marked data belonged to each category in the space was merged and error points were removed by using density clustering algorithm. According to the principle of land leveling, regional division was finished by adjusting the amount of digging earth and the size of the farmland area. Finally, farmland experiments were carried out and the results showed that the proposed method could be used to guide the driving well in the process of land leveling. The sum proportion of overload and empty load during the time when the land was planning and navigation functions were working was no more than 5%, which was better than the situation without these functions. The condition of farmland terrain under each region was improved significantly after being leveled. The value of the standard deviation of altitude was less than 6 cm and it could meet the preliminary goal.