Object Recognition for Millimeter Wave SAR Images Based on Dual-Branch Multiscale Fusion Network

There are several major challenges in the detection and identification of contraband in millimeter wave synthetic aperture radar (SAR) security images: the complexities of small target sizes, partially occluded targets, and overlap between multiple targets are not conducive to the accurate identification of contraband. To solve these problems, a contraband detection method based on the dual branch multiscale fusion network (DBMFnet) is proposed. In the feature extraction stage, the feature extraction network is designed with a dual-branch parallel output structure. One branch preserves high resolution, while the other branch extracts rich semantic information through multiple downsampling operations. Bilateral connections are established between high-resolution and low-resolution branches to facilitate repeated feature exchange, which enhances the ability to detect small and fuzzy targets. In the feature fusion stage, a multiscale fusion module (MSFM) and a context aggregation extraction module (CAEM) are devised; MSFM is utilized to enhance target edge perception, and CAEM is employed to extract contextual information from low-resolution feature maps for enhancing model segmentation performance while reducing computational complexity. The experimental results show that the proposed semantic segmentation method outperforms the existing semantic segmentation methods in mean intersection over union (mIoU).