In China, a flashing green (FG) indication of 3 s followed by a yellow (Y) indication of 3 s is commonly applied to end the green phase at signalized intersections. Stop-line crossing behavior of drivers during such a phase transition period significantly influences safety performance of signalized intersections. The objective of this study is thus to empirically analyze and model drivers’ stop-line crossing time and speed in response to the specific phase transition period of FG and Y. High-resolution trajectories for 1465 vehicles were collected at three rural high-speed intersections with a speed limit of 80 km/h and two urban intersections with a speed limit of 50 km/h in Shanghai. With the vehicle trajectory data, statistical analyses were performed to look into the general characteristics of stop-line crossing time and speed at the two types of intersections. A multinomial logit model and a multiple linear regression model were then developed to predict the stop-line crossing patterns and speeds respectively. It was found that the percentage of stop-line crossings during the Y interval is remarkably higher and the stop-line crossing time is approximately 0.7 s longer at the urban intersections, as compared with the rural intersections. In addition, approaching speed and distance to the stop-line at the onset of FG as well as area type significantly affect the percentages of stop-line crossings during the FG and Y intervals. Vehicle type and stop-line crossing pattern were found to significantly influence the stop-line crossing speed, in addition to the above factors. The red-light-running seems to occur more frequently at the large intersections with a long cycle length.

Abnormal behaviors of train drivers are a critical factor affecting the operational safety of urban rail transit. To achieve automated and efficient detection while meeting practical deployment requirements, this study proposes an end-to-end Temporal Action Detection network based on skeleton data. The network directly uses skeleton sequences as input, integrates a skeleton topology graph tailored to train driver actions for spatiotemporal feature extraction, and employs a non-shared feature propagation design to enhance classification and regression performance. Evaluated on a custom dataset of driver operations (including both standard and abnormal behaviors), the experimental results demonstrate favorable performance with high mean Average Precision (mAP) and strong accuracy. The findings show that the proposed network can accurately localize and classify driver operational behaviors, enabling precise detection of abnormal actions. Furthermore, its low parameter count and minimal storage requirements highlight strong potential for practical deployment in urban rail transit systems.