【摘要】 文中综合考虑了交通安全、效率和流稳定性，分别选取交通冲突率、相对通行效率和速度熵测度非机动车运行状况；采用Yolov5目标检测算法提取车辆运动数据，并利用实测数据建立了多元回归隔离方式设置效果预测模型；根据预测的运行状况，基于熵权法构建了机非隔离方式客观选取方法.以武汉市古田四路、解放大道、解放公园路段为例进行了实例分析，预测模型标定和实例分析结果表明：流量和速度对不同隔离方式下交通安全、效率和流稳定性均有显著影响，改善非机动车交通状况可优先从速度控制入手；三路段运行状况预测值与实测值相比，平均预测精度达94%;古田四路路段建议保持现有机非隔离方式，解放大道与解放公园路段建议更换机非隔离方式为绿化带隔离和隔离栏隔离，交通冲突率将分别减少68.4%、27.9%,速度熵将分别减少0.006、0.137.更多还原

【Abstract】 Considering traffic safety, efficiency and flow stability comprehensively, traffic conflict rate, relative traffic efficiency and speed entropy were selected to measure the operation status of non-motor vehicles. The Yolov5 target detection algorithm was used to extract the vehicle motion data, and the prediction model of multiple regression isolation mode setting effect was established by using the measured data. According to the predicted operation situation, an objective selection method of machine non-isolation mode was constructed based on entropy weight method. Taking Gutian No.4 Road, Jiefang Avenue and Jiefang Park Road in Wuhan as examples, the results of forecasting model calibration and case analysis show that traffic safety, efficiency and flow stability are significantly affected by traffic flow and speed under different isolation modes, and speed control can be given priority to improve non-motor vehicle traffic conditions. Compared with the measured values, the average prediction accuracy of the three sections is 94%. Gutian No.4 Road section suggests to keep the existing non-isolation mode, while Jiefang Avenue and Jiefang Park sections suggest to replace the non-isolation mode with green belt isolation and fence isolation, which will reduce the traffic conflict rate by 68.4% and 27.9% respectively, and the speed entropy by 0.006 and 0.137 respectively.更多还原