Traditional road monitoring methods, including radar, light detection and ranging (LiDAR), and video surveillance, face challenges in accurately and comprehensively identifying unexpected events and potential risks owing to high costs, low accuracy, and environmental complexities. This study proposes an audio-based monitoring method for detecting abnormal road events aimed at uncovering hidden traffic risks. Audio samples collected from various sources, such as FreeSound and Zapslat, are preprocessed using noise reduction, detection, integration, and digital signal conversion via sample quantization. An audio recognition classification model based on the RUSBoost algorithm is constructed and optimized using sequential model-based optimization (SMBO). This optimization addresses issues related to imbalanced samples and features. Optimization increased the accuracy from 68.9% to 96.5% and reduced the error rate to 3.5%. The true positive rate (TPR), true negative rate (TNR), and positive predictive value (PPV) also showed significant improvements, along with notable enhancements in the false positive rate (FPR), F1 score, matthews correlation coefficient (MCC), and kappa (KAP). For key audio events such as slides and crashes, TPR and PPV exceeded 93.2%. The SMBO-RUSBoost model distinguished traffic events from noise by extracting multidimensional features from audio data corresponding to various traffic events. Its classification performance provides a solid foundation for traffic risk assessment and decision-making. The model offers a novel approach for improving road safety.

To describe the lane-changing decision mechanism of vehicles in the bottleneck section of weaving area and provide a lane-changing decision model in the emergency environment, this paper constructed a collaborative lane-changing decision model of vehicles facing bottleneck section based on the micro-trajectory information of vehicles and the human traffic flow model of social force. This model can provide a lane-changing decision method for the sudden bottleneck environment of intelligent network connection. Firstly, based on the characteristics of lane-changing decision of vehicles in the sudden bottleneck section, the vehicle equivalent mass model was constructed to improve the social force model by considering the types of vehicles and drivers. On this basis, the factors driving vehicle lane-changing were described as automatic driving force, repulsive force among vehicles and repulsive force of obstacles, and the collaborative lane-changing decision model was constructed. Then, 832 microscopic trajectory data of lane change decisions were selected and divided into calibration set and verification set. The model was calibrated using genetic algorithm with acceleration as index and Manhattan distance as objective function. The validity of the calibration method was verified based on simulated data and measured data. Finally, this model was compared with the active lane changing decision model in lane changing direction identification, lane changing intention intensity and model prediction error. The results show that the success rate of lane change direction recognition of the proposed model is 92.6%, the output lane change intention intensity is basically consistent with the measured data, and the predicted RMSPE value decreases by 0.825 on average and RE value decreases by 1.379 on average, which are significantly better than those of the active lane change decision model. The research results can provide a theoretical basis for the identification of vehicle lane change intention in the bottleneck section of intelligent network environment and traffic management and control under emergencies.

Low-grade roads often experience frequent roadside intrusions, leading to serious conflicts and disorder. Accurate prediction of the complex traffic-behavior characteristics on such roads is essential for understanding the mechanisms of traffic accidents influenced by roadside intrusions. For this purpose, we collected videos depicting five types of common roadside intrusions on low-grade highways and urban roads. From these videos, we extracted high-resolution vehicle micro-trajectories, and determined the vehicle speeds as they traversed the intrusion area. Then, we identified characteristic sections within the intrusion area, and analyzed the evolution of spatial and temporal characteristics of the vehicle speed. Finally, we established a vehicle speed prediction model using linear, logarithmic and cubic regressions. Notably, the cubic regression model exhibited superior speed prediction performance in the complex scenarios of the intrusion area. The results showed that speed reduction in the intrusion zone of low-grade urban roads is typically higher than that on highways. The deceleration effect is significant for drivers approaching the intrusion source. Additionally, drivers tend to accelerate through the front intrusion zone when their intentions align with those of the intrusion source. However, in scenarios where predicting the behavioral intentions of the intrusion source is challenging, speed may fluctuate to some extent.