An Anti-Overlap Object Feature Weighted Fusion Detector for X-Ray Prohibited Item Detection

Security checkpoints equipped with X-ray scanners play a crucial role in ensuring public safety by inspecting luggage for prohibited items. However, even trained professionals can easily make errors when manually inspecting X-ray images. To assist in manual inspection, direct application of object detection methods to X-rays can lead to false positives and false negatives due to overlapping objects. To address this issue, this paper proposes an object feature weighted fusion detector (OFWF-Detector) for X-ray prohibited item detection. The OFWF-Detector utilizes an object confidence mask selection (OCMS) mechanism to filter overlapping features, improving detection accuracy. Experimental results on the PIX ray dataset demonstrate that the proposed method effectively reduces false positives and false negatives, achieving superior performance compared to traditional methods. Specifically, the OFWF-Detector enhances the detection of small objects in overlapping scenarios by 12.9 % in terms of APs from Tab. I, showcasing its effectiveness in all the challenges posed by overlapping objects in X-ray images. These findings highlight the potential of the OFWF-Detector to enhance the efficiency and accuracy of security screening systems, contributing to improved public safety in high-traffic areas.