A Multi-Radar Track Fusion Methodology Based on Random Forest Regression

This paper introduces a multi-radar track fusion method based on random forest regression and provides an accurate and stable fusion track. The increasing number of aircraft will lead to congested routes, further leading to safety issues. Therefore, an effective track fusion method can accurately locate the aircraft, thereby ensuring the safety of the aircraft in the case of crowded routes. The basic idea of the method proposed in this paper is to select the radar data of a certain track of a certain day to train the model, and predict the position of the aircraft on the next day of the track through the trained model. As a traditional track fusion algorithm, the Kalman filtering has the problem of requiring accurate error estimation, insensitivity to noise, and long calculation time in the case of large data volume. The neural network method that compensates for these shortcomings also has the disadvantage of poor generalization ability in the case of a large amount of noise. The random forest regression model proposed in this paper can overcome the shortcomings of over-fitting in neural network, so it can achieve better prediction results. And through the real data test, the average error is 40m, compared with the neural network method, the result is increased by 50%.