| As the main artery of the city’s transportation network,freeways and expressways provide an efficient and convenient travel service for vehicles traveling through various groups.However,as the demand for interconnection continues to grow,their traffic composition and road network structure tend to be complicated,resulting in more interchange placements within short distances,creating many weaving areas.Compared with other road sections,the operating behavior of vehicles in the weaving areas is more complex,and the frequent lane changes result in a higher proportion of vehicle-to-vehicle interactions,which can easily cause traffic conflicts and induce collisions.The cloverleaf interchange as a common interchange form in the urban road system,whose weaving length is generally shorter,and road safety and traffic congestion status are also more severe.To this end,this study chose the Beihuan Interchange in Chongqing as the research object and obtained a total of 380 minutes of vehicle operation video of the two test weaving areas by utilizing aerial photography from UAVs with high altitude fixation.Subsequently,a study was conducted on the vehicle operating behavior and driving risk in the weaving areas of the cloverleaf interchange.Firstly,a vehicle identification and tracking framework was built based on the target detection algorithm(YOLOx)and the target tracking algorithm(Deep-SORT).Then,a particular dataset of aerial vehicles was created to solve the issue of detecting small target vehicles poorly in aerial view,and the image-slicing processing method was used.After pre-processing the aerial video and the original trajectories,16,364 complete travel trajectories of the test weaving areas were obtained.In addition,the comparison results with the trajectories of the real vehicle tests show that the vehicle trajectories extracted by the algorithm in this thesis have high accuracy in space and time,which meets the requirements of the subsequent research.Secondly,a surrogate safety measure applicable to unrestrained moving vehicles in the weaving areas was calculated based on the trajectory data.The conflict events were identified and classified according to the vehicle orientation.Next,the vehicles’ speed behavior and driving risk characteristics in the weaving areas were thoroughly examined,including the characteristic parameters’ frequency and spatial distribution.Then,the driving risks of vehicles with different operating conditions and traffic states were discussed in detail.The results show lateral conflicts within the weaving areas are more risky than longitudinal rear-end conflicts.The risk is highest at the entrance in the spatial location,and the conflict frequency tends to decay exponentially with the increase of the zone.Vehicles with different operating conditions and traffic states have significant differences in speed and driving risk characteristics.The traveling speed of mainline driving vehicles is greater than that of merging and diverging vehicles.Still,the conflict risk of merging and diverging vehicles is higher.And as the density of the weaving area increases,the risk of conflict also gradually increases.Finally,the influence mechanism of conflict risk in the weaving area was analyzed based on the characteristic parameters of traffic flow,and the risk indicator assessment model was constructed comparing with linear regression,random forest regression,and CatBoost regression.In addition,a prediction model for conflict severity classification was erected.The results show that the developed risk assessment prediction model performs well predicting traffic conflict events.Its effective combination with engineering practice can provide early warning and intervention for dangerous driving behaviors in weaving areas of the cloverleaf interchange.It can also be extended to other road scenarios. |
