Automatic Target Detection using Ground Penetrating Radar (GPR) Data Based on a Machine Learning Approach

Automatic target classification in radar-based systems is critical for various applications, including defense, surveillance, and environmental monitoring. This research presents a robust approach for classifying targets in ground-penetrating radar (GPR) data by combining texture-based Haralick features and shape-oriented Histogram of Oriented Gradients (HOG) features. The integration of these complementary feature sets enhances target characterization by capturing both spatial textures and edge patterns in GPR images. Using the extracted features, we evaluate two prominent machine learning classifiers, Support Vector Machine (SVM) and Random Forest (RF), for their classification accuracy, precision, and computational efficiency. Experimental results on a benchmark GPR dataset show that the fusion of Haralick and HOG features significantly improves classification performance compared to individual feature sets. The SVM classifier achieves superior accuracy with 95.83% in predicting the target/non-target object as mine/non-mine. Further, mine target is detected with image annotation using image morphology. Our method highlights the potential of hybrid feature extraction and machine learning classifiers in achieving accurate and reliable automatic target classification in GPR data, paving the way for real-time applications in complex operational environments. Key Words: Machine Learning, Ground Penetrating Radar, Object Detection, Signal Processing.