Passive Millimeter Wave Concealed Objects Detection Network Using Multi-Polarization Information

Passive millimeter wave (PMMW) imaging offers significant advantages, including the ability to penetrate clothing and the absence of ionizing radiation, making it an effective method for detecting concealed objects on the human body. However, PMMW images are often characterized by a low signal-to-noise ratio (SNR), which poses challenges for target detection. Current PMMW image target detection methods primarily rely on a single polarization, limiting their effectiveness. We introduce a multi-polarization approach to enhance the performance of deep learning network models. Utilizing the efficient target detection framework of YOLO11 as a foundational model, we develop a multi-polarization feature fusion module. This innovative approach facilitates accurate detection of concealed objects within PMMW images. The proposed multi-polarization PMMW image target detection model demonstrates impressive results, achieving a precision of 97.6% and an average precision (AP) of 0.502. These findings indicate that our model significantly improves the detection capabilities of PMMW imaging, addressing the limitations of conventional single-polarization methods.