Concealed Object Detection For Active Millimeter Wave Imaging Based CGAN Data Augmentation

Abstract

Considering under-controlled privacy issues and no health hazards, the active millimeter wave (AMMW) imaging technique has been widely applied in security industries. The ultimate goal is to recognize and detect the concealed object accurately and fleetly, which requires complete and representative datasets. In this paper, concealed object detection for AMMW is proposed. The conditional generative adversarial network (CGAN) is utilized for data augmentation, which enhances the image feature. Data feasibility for detection is validated by the object detection network. Experimental results demonstrate that the proposed method can improve the recognition accuracy effectively and provide a solution for training with small sample datasets.