Deep Learning-Based Detection Using Different Polarization Passive Millimeter-Wave Images

Passive millimeter-wave (PMMW) imaging technology has emerged as a prominent research focus in security screening applications due to its unique advantages including all-weather capability, harmless radiation, and strong material penetration. While polarization represents a fundamental characteristic of electromagnetic waves, current PMMW target detection systems often underutilize this critical information. Meanwhile, deep learning-based approaches for PMMW image analysis have gained significant research attention. This study bridges these two domains by systematically investigating the role of polarimetric information in deep learning-based target detection. Our experimental results, validated through both qualitative visualization and quantitative metrics, demonstrate: significant performance variations across different polarization states, and superior detection accuracy achieved through intelligent multi-polarization fusion strategies.