Parallel Channel Separate Attention Network for Concealed Object Detection in Millimeter-Wave Images

The active millimeter-wave scanner plays an increasingly pivotal role in public safety by employing a non-contact method to detect contraband concealed beneath human clothing. However, millimeter-wave images encounter challenges such as low signal-to-noise ratio, limited resolution, and suspicious targets being small in size, when compared to optical images. To address these challenges and improve detection performance, this paper introduces a Parallel Channel Separate Attention (PCSA) method. Specifically, we propose a parallelized and channel-separated attention module to enhance the extraction ability of features pertaining to minute targets. This module is integrated into the YOLOv8 backbone network, namely PC SA-YOLO. Moreover, we incorporate high-frequency information from the original image into the network input to provide preliminary guidance. Extensive experiments conducted on millimeter-wave datasets demonstrate that our proposed method outperforms existing state-of-the-art detection techniques.