To solve the problems of low accuracy in heading measurement for single-antenna GNSS agricultural machinery navigation system at low speed and the difficulty of starting up, research on agricultural machinery heading measurement technology was conducted based on two-position method and improved STEKF. A suitable heading measurement method for GNSS/IMU fusion in single-antenna agricultural machinery was designed. The heading of high-speed agricultural machinery was measured by the two-position method, and the heading of low-speed agricultural machinery was measured by the improved STEKF algorithm based on the 9-axis data output by the IMU and the position and speed information output by the GNSS. A one-dimensional Kalman filter fusion method based on the two-position method and the improved STEKF heading was designed. The M-1204 tractor produced by Lovol was used as the experimental platform to verify the heading measurement accuracy and stability under speed gradient change conditions. The results showed that during the process of starting up from a stationary state and traversing mixed routes containing straight lines and curves at a speed of 0.36km/h to 5.40km/h, the heading angle obtained by the proposed fusion heading measurement method was within 0.5° of the heading angle obtained by the dual-antenna GNSS, indicating that the proposed method was suitable for stable heading acquisition of agricultural machinery in the state of starting up and low-speed movement, which improved the precision and stability of single-antenna agricultural machinery navigation and low-speed driving control and provided technical basis for high-precision agricultural machinery autonomous unmanned operation.