为科学评价信号交叉口右转渠化岛设置给非机动车过街带来的交通安全风险，实地采集了南昌市5个信号交叉口的视频，并提取了各交叉口右转渠化岛区域机非冲突数据、交叉口几何设计特征、交叉口渠化方式以及交通流特性等数据；采用改进型距离碰撞时间（time to collision, TTC）对交通冲突进行判定，共得到304起冲突事件；基于贝叶斯方法构建了考虑不同交叉口间异质性的随机效应冲突模型，对比了固定效应与随机效应负二项模型的拟合优度并分析了显著影响因素；基于随机效应模型，确定了期望机非冲突数计算公式，绘制了不同交通流情况下信号交叉口渠化岛设置标准曲线并提供了案例说明。研究表明：相比于固定效应模型，随机效应模型对机非冲突有更好的拟合效果；相比于无渠化岛交叉口，软渠化岛和硬渠化岛的设置会导致交通冲突分别增加38%和61.4%；右转机动车交通量、过街非机动车交通量和非机动车违法行驶数量每增加1%，将导致机非冲突分别增加0.85%、0.44%和0.18%；右转机动车平均速度每增加1%，将导致机非冲突减少2.5%。研究成果为定量化分析交通冲突影响要素提供了有效思路，可为右转渠化岛交通设计提供理论支撑。

In order to scientifically evaluate the traffic safety risk to non-motorized vehicles caused by the setting of right-turn channelized island at signalized intersections, the video of five signalized intersections in Nanchang was collected in the field, and the conflicts data between motor and non-motor vehicle in the area of right-turn channelized island at each intersection, geometric design characteristics of intersections, intersection channelization treatment and traffic flow characteristics were extracted. The improved TTC (time to collision) index was used to determine the traffic conflicts, and a total of 304 traffic conflicts were obtained. Based on the Bayesian method, a random-effect conflict model was constructed to take into account the heterogeneity of different intersections. The fit of the random-effect and fixed-effect negative binomial models was compared and the significant influencing factors were analyzed. Based on the random-effect model, the formula for calculating the number of expected conflicts between motor and non-motor vehicle was determined and the standard curves for the setting of channelized island were plotted under different traffic conditions and a case description was provided. It is concluded that the random-effects model has a better fit for conflicts between motor and non-motor vehicle than the fixed-effects model. Compared with the intersection without channelized islands, the installation of paint and raised channelized islands will increase traffic conflicts by 38% and 61.4% respectively, each 1% increase in the volume of right-turning motor vehicle traffic, the volume of crossing non-motor vehicle traffic and the number of illegal non-motor vehicle movements will lead to an increase in conflicts between motor and non-motor vehicle by 0.85%, 0.44% and 0.18% respectively. A 1% increase in the average speed of right-turning motor vehicles would result in a 2.5% reduction in the conflicts. The research results provide effective ideas for quantitatively analyzing the impact elements of traffic conflicts, which can provide theoretical support for the traffic design of right-turn channelized islands.