摘要
为提高客流预测的精度,构建轨道交通站点客流多变量时间序列预测模型。基于视频检测的轨道交通短时客流预测研究采用方向梯度直方图特征描述器与支持向量机分类器识别行人目标,利用Camshift算法对目标跟踪,从而获取客流量和客流速度参数,并根据协整关系构建客流多变量预测的向量误差修正模型,最后利用南京鼓楼车站4A通道的视频数据进行模型验证和对比分析。实例验证结果表明:构建的向量误差修正模型具有较好的预测性能,客流量和速度预测的MAPE值都小于8%,优于相同样本下ARIMA(0,1,1)的预测性能。
In order to improve the accuracy of passenger flow state prediction, a multivariate time series forecasting model is proposed for the forecast of the passenger flow state. For the research on short-time passenger flow forecast based on video frequency, it adopts the directional gradient histogram feature descriptor and support vector machine detection to identify pedestrian targets, and uses the Camshift algorithm to track the target, so as to access the parameters of passenger flow and speed, and then establish the vector error correction model according to the co-integration relation between parameters. Finally, validation and comparative analysis are carried out using 4A channel video frequency data measured at Gulou station in Nanjing, the results show that the model constructed in this paper has a better prediction performance. The MAPE value of both the passenger flow and speed is less than 8%, which is better than the ARIMA （0,1,1） prediction performance established by the same sample data.
出处
《铁路通信信号工程技术》
2018年第1期50-55,共6页
Railway Signalling & Communication Engineering
基金
交通运输部建设科技项目(2015318J33080)
江苏省重点研发计划(社会发展)项目(BE2016740)
南京地铁运营有限责任公司专项项目(8550140226)
关键词
城市轨道交通
短时预测
向量误差修正模型
性能评估
urban rail transit
short-time prediction
vector error correction model
performance evaluation
