基于多源数据融合的多层次轨道交通网络客流量监测方法

A multi-level passenger flow monitoring method for rail transit networks based on multi-source data fusion

在智能交通领域,客流量监测是一个研究热点,所获数据对于交通资源的适时调度与合理配置等有着重要的指导意义。现有的轨道交通客流量监测方法存在预测性较强,监测结果不够准确的问题,设计基于多源数据融合的多层次轨道交通网络客流量监测方法,实现较大客流量下的轨道交通网络客流量监测。在多层次轨道交通网络中大量布设高清网络摄像机与红外线摄像机以采集客流量监测图像。对于所采集的两种外图像,设计基于图信号稀疏表示的图像去噪算法,对其实施去噪处理。设计基于DPM模型的行人检测算法,实施可见光图像与红外图像的行人运动目标检测。对于两种图像的检测结果,设计基于密集交叉网络的多源数据融合算法,实施多源数据融合与计数,实现多层次轨道交通网络客流量监测的目的。测试结果表明,该方法的高铁网、城际轨迹网、市郊铁路网、城市轨道网客流量监测误差分别为44人、40人、19人、4人,客流量监测误差均较低。

In the field of intelligent transportation,passenger flow monitoring is a research hotspot,and the data obtained has important guiding significance for timely scheduling and reasonable allocation of transportation resources.The existing methods for monitoring rail transit passenger flow have the problem of strong predictability and inaccurate monitoring results.A multi-level rail transit network passenger flow monitoring method based on multi-source data fusion is designed to achieve monitoring of rail transit network passenger flow under large passenger flow.A large number of high-definition network cameras and infrared cameras are deployed in multi-level rail transit networks to collect passenger flow monitoring images.Design an image denoising algorithm based on sparse representation of image signals for the two collected external images,and apply denoising processing to them.Design a pedestrian detection algorithm based on the DPM model,and implement pedestrian motion target detection in visible and infrared images.For the detection results of two types of images,design a multi-source data fusion algorithm based on dense cross network,implement multi-source data fusion and counting,and achieve the purpose of multi-level rail transit network passenger flow monitoring.The test results show that the passenger flow monitoring errors of the high-speed rail network,intercity trajectory network,suburban railway network,and urban rail network of this method are 44 people,40 people,19 people,and 4 people,respectively,with low passenger flow monitoring errors.