A hybrid model of an object detector on synthetic aperture radar images with the aim of enhancing the utilization of radar shadow information

This investigation presents an advanced hybrid model for object detection in Synthetic Aperture Radar (SAR) images, focusing on improving the use of radar shadow information to reduce false alarms and enhance detection accuracy. The proposed model is based on the intelligent fusion of two complementary approaches: one specializes in detection using shadow information, while the other focuses on detection independent of shadows. The selection of the appropriate model or the fusion of their outputs is carried out through an adaptive decision-making mechanism based on image quality analysis, including calculating noise levels using Fourier transform and a blur index using textural features (GLCM). The results demonstrated a clear superiority of the hybrid model, which achieved a significant reduction in the false alarm rate (FPPI) to 0.0035 on the original test set and to 0.0061 on the composite set at a confidence threshold of 0.8, while maintaining high recall and overall accuracy. These results make the hybrid model a practical and promising solution for military and security applications requiring high reliability in cluttered and camouflaged environments.