REAL-TIME AIR-BORNE TARGET CLASSIFICATION USING KINEMATICS DATA FOR COASTAL SURVEILLANCE RADAR

Abstract

Aircraft target classification is an important component of modern defense and surveillance systems. This paper presents a machine learning-based framework for real-time aircraft classification using kinematic trajectory parameters. The proposed system utilizes statistical and temporal feature extraction methods combined with supervised learning algorithms. Experimental evaluation demonstrates strong classification accuracy and real-time operational capability suitable for coastal surveillance radar systems. The system processes real-time trajectory data containing parameters such as height, resultant velocity, resultant acceleration, and Automatic Gain Control (AGC) values received through UDP-based communication. A sliding window mechanism is employed to extract meaningful statistical and temporal features from continuous time-series data. Multiple machine learning algorithms including Logistic Regression, K-Nearest Neighbors, Support Vector Machine, Decision Tree, Random Forest, Gradient Boosting, and XGBoost were trained and comparatively evaluated. Among all models, the Random Forest classifier achieved the best performance with high classification accuracy and robustness. The proposed framework also integrates an interactive visualization dashboard for real-time monitoring, analytics, and prediction reporting. The developed
system demonstrates low-latency prediction capability, making it suitable for operational deployment in defense monitoring and coastal surveillance applications. Furthermore, the proposed methodology provides scalability, reliability, and improved decision-making support for intelligent real-time airborne target recognition systems for coastal region.