In order to improve the real-time measurement accuracy of field surface topography in the area behind the leveling in the process of farmland leveling operation, a real-time measurement method of 3D topography in the process of farmland precision leveling was proposed.Using a supported-wheel dryland grader installed with GNSS dual antennas and attitude sensors as a terrain measurement platform, the Kalman filter was firstly used to fuse GNSS and acceleration to improve the positioning accuracy and frequency, and then the vehicle coordinates of the bottom point of the supported wheel were obtained by establishing the kinematic model of the grader, and finally the geodetic coordinates of the bottom point of the supported wheel were solved by combining the information of the position of grader and the topographic map was generated by using the nearest neighbor interpolation method.The test verification showed that this method can accurately solve the coordinates of the bottom point of the support wheel, and the root mean square error of the plane measurement was less than 10mm, and the root mean square error of the elevation deviation was not more than 20mm.The test results of cement pavement showed that the root mean square error of the elevation difference between the measured value and the true value was less than 30mm when the three-dimensional terrain of the same section of cement pavement was measured at three different speeds.The field test results showed that the root mean square error of the measured terrain was 16.5mm,flatness was 16mm, and the elevation difference distribution less than 30mm was 95.8%. Compared with the airborne GNSS measurement method, the root mean square error accuracy was improved by 29.5%, flatness accuracy was improved by 11.1%, and height difference distribution column accuracy was improved by 9.5%.The method proposed can obtain the topographic information of the area behind the leveling in real time and accurately during the leveling operation, and provide a basis for the research of real-time path planning of unmanned farmland leveling.