AutoML-Based Fusion of Radar and Image Data for Aerial Target Classification

This study aims to detect and classify aerial targets by integrating radar data with image processing techniques. While radar systems provide essential physical parameters such as target position and velocity, image processing methods including grayscale conversion and feature extraction enable the analysis of visual characteristics. This multimodal integration significantly enhances classification performance, particularly in distinguishing between various aerial platforms such as unmanned aerial vehicles (UAVs), fighter jets, and helicopters. Rather than acting as a stand alone classifier, FLAML (Fast and Lightweight AutoML) an automated machine learning (AutoML) framework that automatically evaluates and optimizes a range of candidate machine learning models was employed to identify the most suitable classifier for the task. Among these, ensemble based models namely XGBoost (Extreme Gradient Boosting), LightGBM (Light Gradient Boosting Machine), and Random Forest demonstrated superior performance compared to other approaches. Results obtained from five fold cross validation indicate that both LightGBM and XGBoost achieved comparable and state of the art generalization performance on the test data, with an accuracy and macro F1-score of approximately 84%. These findings reveal that the combined use of radar and image-based data provides an effective and reliable approach for aerial target detection and classification.