X-ray safety inspection machines are essential for maintaining security. In the past few years, the rapid development and widespread application of deep learning have significantly contributed to the progress of X-ray security detection. In order to achieve automatic detection of dangerous objects in X-ray baggage images, this paper proposes an efficient dangerous objects detection network called X-YOLO, which incorporates feature fusion and attention mechanisms specifically for Xray baggage images. In the proposed network, an innovative attention mechanism module is integrated into the architecture, an improved dynamic head module is designed to enhance object detection across various dimensions. In the experiments, the HiXray and OPIXray datasets are selected, and the experimental results show that the X-YOLO network demonstrates excellent performance when compared with several state-of-the-art networks. The mAP50 value reached 86.6% on the HiXray dataset and 93.7% on the OPIXray dataset, representing improvements of 2.4% on the HiXray dataset and 4.6% on the OPIXray dataset compared with the baseline
Traffic light detection technology can assist drivers in making decisions and has potential applications in autonomous driving to reduce loss of life and property. However, the detection of traffic light objects is challenging due to traffic light’s small size and complex backgrounds, making it difficult for current technology to meet real-time and high accuracy requirements. In this work, we identify that inappropriate feature integration and static parameters during detector training limit model efficiency. Specifically, using the same detection criteria for all object sizes in the feature integration process reduces the effectiveness on small objects such as traffic light. Additionally, a static IoU threshold during training results in insensitivity to objects in complex backgrounds. Finally, static parameters in the loss function hinder the learning of high-quality samples. To address these problems, we propose Dynamic-TLD. This approach improves the training efficiency of small samples through the weighted integration of low-level and high-level features. It also enhances the learning of traffic light images by gradually increasing the IoU threshold based on model learning. Additionally, the learning of high-quality samples is facilitated by dynamic parameter selection in the loss function. Compared with six other advanced detectors, our approach improves average precision on three publicly available traffic light datasets by 1.4 points, 2.1 points, and 0.8 points respectively while meeting the real-time requirement of autonomous driving systems. These results demonstrate the effectiveness and robustness of our approach and its potential for application in autonomous driving systems. Codes and demonstration video are available at https://github.com/ljyw17/traffic-light-detector.
Automated threat detection in transported goods using X-ray screening plays a critical role in ensuring the safety and efficiency of modern transportation systems, including highway checkpoints, ports, and airports. Manual inspection, still widely used, struggles with limited accuracy and scalability in high-throughput scenarios. Although existing object detection models perform well on natural images, they often fail to generalize to X-ray imagery due to occlusion, clutter, and the presence of small or non-metallic items. To address these challenges, we propose iX-Det, a novel X-ray threat detection network tailored for transportation security applications. iX-Det integrates two key components: a distraction-aware and path-augmented feature pyramid network (DAPA-FPN), and a C3 attention mechanism (C3-AM). DAPA-FPN enhances multiscale object representation while reducing clutter common in X-ray scans. Meanwhile, C3-AM introduces spatially adaptive attention into the backbone to improve feature extraction for restricted item detection. Extensive experiments on public X-ray datasets demonstrate that iX-Det significantly outperforms existing methods in both detection accuracy and inference speed. These results highlight its potential for deployment in real-world threat screening systems within intelligent transportation infrastructure.