Imputation Of Road Network Traffic Conditions Based On Spatial-temporal Correlation

The real time and complete data of traffic flow is an important factor affecting the effectiveness of intelligent transportation system.and is the basis of network traffic state estimation,short-term traffic flow prediction,traffic guidance,traffic management and control services.However,due to the limitation of collection equipment,the impact of traffic environment and system calculation error,the collected traffic flow data is often invalid or lost,which affects the traffic system and services based on real-time traffic flow data.Therefore,so it is necessary to carry out real-time interpolation for the missing data of urban road network traffic flow.According to the theory of spatial-temporal correlation,the adjacent road segments may have similar traffic flow fluctuations.At the same time,the traffic flow has the periodic law of taking the day or week as the time unit,so the traffic flow of the road network has a certain time and space correlation.By quantifying the degree of spatio-temporal correlation between road networks,we can use the traffic flow data of other relevant sections or periods to estimate the value of the current missing data.This paper,estimate the average speed of the road section by the GPS track data of taxi;constructe the adjacent attribute matrix of the road section based on the topological relationship of the road network,and use Pearson function to quantify the cross-correlation of the traffic flow between the adjacent road sections and the auto-correlation between the traffic flow sequence and the historical data.The correlation coefficient is used as the influence weight of the adjacent road section data and the historical statistical data to impute the missing position,so that we can build the Interpolation model based on spatial-temporal correlation.Considering the sparsity and uneven spatial-temporal distribution of GPS track data of taxi,this paper proposes a combination of real-time correlation coefficient and statistical correlation coefficient,using the correlation coefficient of multi day data statistics as a substitute when the data is losting too much data and cannot calculate the correlation coefficient,so as to adapt to the situation of sparse data in some periods.In this paper,we use the GPS track data of taxi in Shenzhen to carry out experiments on road networks in Nanshan District.The correlation interpolation model of adjacent road sections,historical sequence correlation interpolation model and spatiotemporal correlation interpolation model are selected for experimental comparison.The results show that the spatiotemporal correlation interpolation model,which integrates the temporal and spatial correlation,can achieve 97.46% coverage and 7.27km/h average error after the interpolation of the road network with an average data coverage of 65.41%.It is better than the other two models in interpolation accuracy and data completion rate.And is verified that this method is feasible and effective for sparse traffic flow data interpolation of road network.