<p>Based on millimeter-wave (MMW) images, concealed object detection technology has become an ideal choice for high-throughput security scenarios such as large-scale venues, railway stations, and airports due to its non-contact, real-time detection characteristics. Current model designs primarily focus on whether objects are detected, while neglecting the categories of the objects, which makes it difficult to meet the practical security needs for accurately identifying objects of different threat levels. To address this issue, we propose a Dual Receptive Field Enhancement YOLO Network (DRFE-YOLO). By integrating the spatial adaptive receptive field modulation mechanism of Switchable Atrous Convolution and the channel-wise global receptive field aggregation mechanism of the Squeeze-Aggregated Excitation module, a dual receptive field enhancement system is constructed, which synergizes spatial and channel dimensions to improve the categorical discrimination capability for concealed objects of varying sizes and shapes. To validate the effectiveness and feasibility of the proposed method, we constructed a MMW image dataset containing multiple types of concealed objects and evaluated the model’s performance through comprehensive experiments. Experimental results show that the proposed DRFE-YOLO model achieves significant improvements in mean Average Precision (mAP50 and mAP50-95) compared to the YOLOv8n baseline model, with increases of 8.7% and 5.9%, respectively. Furthermore, the model demonstrates enhanced detection performance across all object categories, particularly showing significant improvement in detecting small targets such as knives and lighters. With excellent performance, this model is expected to advance the practical application capabilities of MMW security scanners in the field of public safety.</p>