Object Segmentation Using Polarization Random Feature in Passive Millimeter-Wave Imaging

Object segmentation is an important issue in the field of passive millimeter-wave imaging remote sensing and detection. The brightness temperature (TB) difference of millimeter-wave radiation is usually used for the segmentation and detection of different target types. However, the target TB is affected by many factors such as dielectric constant, shape, physical temperature, and environmental radiation. In the actual scene, it is difficult to achieve effective object segmentation only depending on TB difference. In this paper, a physically-based method for object segmentation based on polarization random features is proposed. Through an in-depth analysis of the physical model and characteristics of the angle of polarization (AoP), the AoP statistical distributions of various polarized and non-polarized targets were given. We found that the AoP random feature is very sensitive to the polarization characteristics of the object, and its local statistical standard deviation can be used for polarized and unpolarized object segmentation. Two multi-polarization imaging experiments of complex scenes have verified the effectiveness of the proposed method. Compared with several methods based on polarization degree, the superior performance of the proposed method is qualitatively and quantitatively verified. Our work breaks through the inherent thinking that AoP is generally used for three-dimensional reconstruction, and opens up a new perspective of object segmentation based on AoP.