Intelligent Identification of Airborne Target Trajectory Behaviour Based on Long Time-Series Space-Based Radar Data

As the intensity of confrontation in the neighbourhood continues to escalate, the air defence threat facing our country is growing, and information-based air defence is facing unprecedented challenges. On the basis of existing aviation detection means, comprehensive analysis of airborne targets is the primary task of exploring the laws of information-based air defence. Among them, the civil-military identification of flying targets is crucial to the military's operational decision-making and battlefield control. Hence, this study focuses on trajectory signal processing, machine learning intelligent recognition based on manual sequence features, and improvement of TSFRESH feature extraction to enhance its timeliness. Specifically, we design a rule-based filtering method to effectively remove wild-value data from the original trajectory data. And then we propose a support vector machine model based on manual sequence features. Effective trajectory features are extracted based on TSFRESH, and the support vector machine model is used to achieve the classification of military and civilian aircraft. Meanwhile to shorten the TSFRESH feature extraction time, we use the mutual information method to downscale the extracted features to speed up the classification time. Experiments show that this method can achieve the task of classifying military and civilian aircraft based on trajectory information and significantly reduce the running time.