Aiming to address the issue of insufficient positioning accuracy of ultra-wideband (UWB) positioning technology in agricultural greenhouse environments, caused by poor interference immunity and unknown statistical characteristics, a UWB positioning technology was proposed-based on an improved adaptive Kalman filter (IAKF) algorithm. Firstly, an anomaly detection mechanism was introduced to identify divergence phenomena during the filtering process. Subsequently, the measurement noise covariance matrix was updated in real-time to suppress filter divergence and enhance the algorithm’s adaptability in the presence of strong noise fluctuations. Simulation positioning experiments under three different noise environments were conducted to compare and analyze the performance of UWB, IAKF, adaptive Kalman filter (AKF), and Kalman filter (KF) algorithms. The simulation results showed that the IAKF algorithm exhibited stronger adaptability and robustness. Finally, using a self-developed agricultural tracked vehicle as the positioning carrier, UWB positioning experiments were conducted in the greenhouse environment. The experimental results indicated that in the greenhouse environment, the positioning accuracy of the tracked vehicle using the IAKF algorithm was improved by 22.2% and 13.0% in-line of sight (LOS) and 20.0% and 15.4% in non-line of sight (NLOS) scenarios compared with that of the AKF and KF algorithms, respectively.