查看文章

Millimeter wave (MMW) images can be used to identify metallic or non-metallic embargoed items hidden under clothing in both active and passive modes. Detection of concealed objects aims to locate these forbidden items from given MMW images. However, how to effectively realize the detection of concealed objects from MMW images is still a very challenging task, since the prohibited items are relatively smaller in size than human bodies, and the quality of MMW images is also relatively lower than natural images such as in signal-to-noise ratio and spatial resolution, and further, domain shifts of MMW images are common in the real applications. In this paper, a domain adaptive detection system, mainly composed of two parts: a detection module and a domain adaptation module, is proposed. Our system takes reconstructed single-channel MMW images as the input. For the detection module, a top-down …