基于小波卡尔曼滤波的高速公路交通数据融合去噪算法研究

Research of Highway Traffic Data Fusion De-noising Algorithm Based on Wavelet-kalman Filter

为去除交通信息采集过程中的噪声干扰,提出了一种基于小波分析和卡尔曼滤波相融合的交通数据去噪算法。该算法通过小波系数计算小波方差并代替卡尔曼滤波的初始协方差完成迭代,将小波阈值去噪重构后的信号作为卡尔曼滤波器状态最优估计中的测量值输入,实现了交通数据的分解去噪和最优估计。实例分析结果表明:一方面小波-卡尔曼滤波融合去噪算法的去噪指标优于小波分析算法;另一方面采用去噪后的实时交通数据建立时间序列预测模型,由三项预测误差评价指标及拟合预测图对比可知,小波-卡尔曼滤波融合去噪算法较小波分析算法可更好地提高预测精度,从而综合验证该融合算法能有效去除交通数据中的噪声干扰,并提高其数据质量。

In order to remove the noise interference in the process of traffic information collection,a fusion de-noising algorithm based on wavelet analysis and Kalman filter was proposed.The algorithm calculated the wavelet variance by wavelet coefficients,and the initial covariance of Kalman filter was replaced by wavelet variance to complete the iteration;The signal reconstructed by wavelet threshold de-noising was input as the measured value in the Kalman filter state optimal estimation,the decomposition de-noising and optimal estimation of traffic data were realized.The research results show that on the one hand,the de-noising indexes of Wavelet-kalman fusion de-noising algorithm are better than that of wavelet analysis algorithm;on the other hand,the time series forecasting models are established based on de-noising real-time traffic data,by comparing the three predictive error evaluation indexes and the fitting prediction graph,the Wavelet-Kalman fusion de-algorithm is better to improve the prediction accuracy than that of wavelet analysis algorithm,the above two aspects integrated verify that the fusion algorithm can effectively remove noise in traffic data,and improve the quality of traffic data.