基于航拍视频的交叉口车流量统计方法研究

Research on Statistical Method of Traffic Flow at Intersection Based on Aerial Video

针对现有的智能车流量统计方法只局限于单一方向路段检测的问题,采用自适应混合高斯模型和数学形态学运算来分离航拍视频的背景和前景,提出一种适用于交叉口复杂场景的改进的组合虚拟检测线方法,按照位置坐标分别在进口道和出口道位置设置检测线,检测线两两组合成对儿,并用投影法确定统计阈值关系,然后统计交叉口四个方向的车辆不同转向的交通流量。以唐山市学院路与长宁道交叉口和建设路与南新道交叉口的航拍视频为例,采用所提出方法统计交叉口的车流量,最后得到的误检率低至4.67%和4.41%,证明本文方法的具有良好的可行性。

Aiming at the problem that the existing methods of intelligent traffic flow statistics are limited to the de-tection of single-direction road section,adaptive mixture Gaussian model and mathematical morphology operation were used to separate background and foreground of aerial video,and an improved combined virtual detection line method was proposed in this paper,which is suitable for the complex conditions of intersections.According to the lo-cation coordinates,the detection lines were set at the entrance and exit lanes respectively,and the detection lines were paired in pairs.The statistical threshold relation was determined by the method of projection,then the traffic flow of vehicles in different directions of the intersection was counted.Taking the aerial videos of two intersections of Xueyuan Road and Changning Road,Jianshe Road and Nanxin Road in Tangshan City as an example,the traffic flow at intersections was calculated by using the method proposed in this paper,and the final false error detection rate is as low as 4.67%and 4.41%.The feasibility of this method is demonstrated.