Speed illusion is the leading contributing factor to traffic accidents in highway tunnels. This study aimed to estimate the influence of visual information at different scales and frequencies on drivers’ visual perception and driving safety in highway tunnels. The speed perception of drivers was measured using the stimulus of subjectively equivalent speeds as an index. Thirty drivers were recruited to conduct a psychophysical experiment on speed perception using a driving simulator. The large-, medium- and small-scale visual information in a frequency range of 0.1–32 Hz were used in the experimental scene to generate scenes for comparison. The results show that high-frequency visual information (2–32 Hz) might lead to driver overestimation of vehicle speed in tunnels, while medium-frequency (0.4–1 Hz) and low-frequency (0.1–0.2 Hz) visual information contribute to speed underestimation. The medium-scale information had the largest speed overestimation effect, followed by large- and small-scale information (significant differences of 2–8 Hz). Medium-scale visual information below 8 Hz had the lowest degree of dispersion of speed perception. Therefore, the use of integrated high-frequency, medium-scale visual information and medium-frequency, large- and small-scale visual information is suggested to reduce the speed illusion of drivers and ensure driving safety.

Many small-spacing interchanges (SSI) appear when the density of the expressway interchanges increases. However, the characteristics of traffic accidents in SSI have not been explained clearly. Therefore, this paper systematically takes the G3001 expressway in Xi’an as the research object to explore the accident characteristics of SSI. Firstly, the expressway is divided into four sections. Furthermore, their safety can be evaluated by the number of accidents per unit distance of 100 million vehicles (NAP). Subsequently, eight indexes, such as mean spacing distance (MSD), are selected to explain the cause affecting expressway safety by developing the least square support vector machine (LSSVM). Secondly, the difference between SSI and normal-spacing interchanges (NSI) is clarified by statistical analysis. Finally, LSSVM, random forest, and logistic regression models are built using 12 indicators, such as the time spent exploring the causes of serious accidents. The results show that the inner ring NAP in Sections I and II with SSI is 27.2 and 33.7, higher than in other sections. The density, annual average daily traffic, and MSD adversely affect expressway traffic safety. The road condition mainly influences the serious traffic accidents in the SSI. This study can provide the theoretical basis for traffic management and accident prevention in the SSI of the expressway.

Urban expressway weaving sections suffer from a high crash risk in urban transportation systems. Studying driving behavior is an important approach to solve safety and efficiency issues at expressway weaving sections. This study aimed to investigate the influence of drivers’ individual differences on diverging behavior at expressway weaving sections. First, a k-means cluster analysis of 650 questionnaires was performed, to classify drivers into three categories: aggressive, conservative and normal. Then, the driving behavior of 45 drivers from the three categories was recorded in a driving simulator and analyzed by an analysis of variance. The results show that different types of drivers have different driving behaviors at weaving sections. Aggressive drivers have a higher mean speed and mean longitudinal deceleration, followed by normal and conservative drivers. Significant differences in the range of lane-change positions were found between 100, 150 and 200 m of weaving length for the same type of drivers, and the duration of weaving for aggressive drivers was significantly smaller than for normal and conservative drivers. A significant correlation was found between lane-change position and weaving duration. These results can help traffic engineers to propose effective control strategies for different types of drivers, to improve the safety of weaving sections.

Vehicle detection and tracking from unmanned aerial vehicles (UAVs) aerial images are among the main tasks of intelligent traffic systems. Especially in tasks with long distances, extensive backgrounds, and small objects, it increases the difficulty of localization and regression, which can easily lead to missed detections and false positives. This paper proposes a detection-based small-scale vehicle tracking framework that integrates an improved YOLOX network and the DeepSORT algorithm to address these issues. Based on the original YOLOX network, a shallow feature extraction network, 160 × 160 pixels, is added to enhance the ability to extract small-scale object features. A convolutional block attention module (CBAM) is inserted in front of the neck network to select crucial information for vehicle detection tasks while suppressing noncritical ones. EIoU_Loss is introduced as the bounding box regression loss function in training to speed up their convergence and improve the localization accuracy of the small objects. Furthermore, an image segmentation method is proposed to effectively reduce missed and false detection events. It divides the original high-definition image into multiple subimages, first detected and then reassembled. Finally, the improved YOLOX network is used as the detector of the DeepSORT to perform small-scale vehicle detection and tracking tasks in various traffic scenarios. Experiments show that the proposed method can significantly improve the detection accuracy of the network and effectively solve the problems of missed detection and false positives in small-scale vehicle tracking tasks in high-resolution aerial images captured by high-altitude UAVs. Significantly, the algorithm proposed in this paper has sufficient robustness for small-scale tracking tasks of aerial videos captured at different altitudes.