Passive millimeter-wave (PMMW) is a significant potential technique for human security screening. Several popular object detection networks have been used for PMMW images. However, restricted by the low resolution and high noise of PMMW images, PMMW hidden object detection based on deep learning usually suffers from low accuracy and low classification confidence. To tackle the above problems, this paper proposes a Task-Aligned Detection Transformer network, named PMMW-DETR. In the first stage, a Denoising Coarse-to-Fine Transformer (DCFT) backbone is designed to extract long- and short-range features in the different scales. In the second stage, we propose the Query Selection module to introduce learned spatial features into the network as prior knowledge, which enhances the semantic perception capability of the network. In the third stage, aiming to improve the classification performance, we perform a Task-Aligned Dual-Head block that decouples the classification and regression tasks, and enhances the interaction between the two heads. Based on our self-developed PMMW security screening dataset, experimental results including comparison with State-Of-The-Art (SOTA) methods and ablation study demonstrate that the PMMW-DETR obtains higher accuracy and classification confidence than previous works, and exhibits robustness to the PMMW images of low quality.

Passive millimeter wave (PMMW) imagers are increasingly becoming practical sensor candidates for target detection tasks. This paper is devoted to the ship detection by land-based W band passive polarized imager. The radiation characteristics of sky, sea surface and the atmospheric absorption characteristics of sea area near Qingdao are calculated based on sounding data. And the W band radiation characteristic of painted metal is measured. Then, the experiments of detection for sea surface ship are carried out in different ranges including 2,5, and 15 kilometers. The results show that land-based PMMW imager can detect and locate ship target from sea surface more than tens of kilometers away in thick fog, and polarization selection has influence on detection performance.

Passive millimeter-wave imaging has emerged as a useful tool in many remote sensing applications, including resource re-mote sensing, material classification and target detection. This paper presents a method to classify specular objects based on their material composition from passive millimeter-wave polarimetric imagery. Passive Degree of Polarization (PDoP) is proposed and cal-culated by using the results of passive millimeter-wave polarization measurement. The PDoP values of typical ground targets are further analyzed. Outdoor experiments are conducted, and the PDoP image is generated from the brightness temperature images. The PDoP values of pixels in the image are statistically analyzed, and the threshold values are set based on the statistical analysis results, then every pixel is recognized. Experimental results indicate that this method is highly effective for distinguishing among various materials of interest.

Passive millimeter-wave (PMMW) sensing has been used for several close-range applications such as scene monitoring and security check. Aiming to sense man-made or natural dielectrics, we report the estimation characteristics of complex permittivity by multi-polarization measurements. According to the polarization brightness temperature (TB) model, the complex permittivities of dielectric materials are estimated by using the TB data of multiple linear polarizations. By Monte Carlo simulations, we found that the estimated complex permittivities can form a special cluster. All estimation values of each dielectric spread around a special nonlinear curve (called the “C-curve”). Then the “C-curve” distribution characteristics of different parameter conditions such as polarization number, incident angle, roughness, and measurement noise, have been analyzed. The 94 GHz multi-polarization measurements have been conducted to prove the existence of “C-curve” feature. The results suggest that this invariant feature can be possibly used for material classification and image segmentation in PMMW sensing.

Surface normal vector is essential for object recognition and three-dimensional reconstruction. This article establishes a passive millimeter-wave polarization observation model of object thermal radiation. Then, a polarization-based method is proposed to acquire the surface normal vector from multipolarization measurements. The simulation results show that our method has high accuracy and robustness. The outdoor imaging experiment was conducted, and the polarization characteristics of several typical objects in the imaging were analyzed. Experimental results show that our method can successfully estimate the surface normal vector. The possible applications of our method include object recognition and pavement inspection.

Millimeter-wave (MMW) radiation characteristic has been used in object classification and recognition. Complex permittivity is a key factor affecting the radiation characteristic. Previous study has shown that estimation values of complex permittivity based on multipolarization measurements spread around a special curve (called ``C-curve''). In this letter, we analyzed the influence of complex permittivity on MMW radiation. We find that ``C-curve'' is the manifestation of the solution instability and can be avoided by our new estimation method based on the measurements of multiple incident angles. The simulation and experiment have proved the validity of our method. During research, we also find that objects whose complex permittivities are near the ``C-curve'' are hard to distinguish by MMW radiometry, unless we measure their radiation of vertical polarization near the Brewster angle. This finding has the guiding significance for object classification, recognition, and information acquisition.