Crowdsensing Based Traffic Velocity Missing Data Recovery Algorithm
ZHANG Jianzong，TAO Dan
（Institute of Signal Processing and Artificial Intelligent，School of Electronic and Information Engineering，Beijing Jiaotong University，Beijing 100044，China）
Abstract：The popularization of GPS makes it possible to evaluate the condition of urban roads based on largescale vehicle data.In this paper，we study the problem of traffic velocity missing data recovery based on crowdsensing.Firstly，vehicle data is collected by probe vehicles，and a grid density based road network extraction method is designed.Secondly，according to the characteristics of GPS data，a selfadaptive road segment velocity calculation method is designed in order to obtain the traffic velocity matrix.Thirdly，we classify the missing data in traffic condition evaluation，and propose an improved sparse representation on basis of compressed sensing by considering the spatiotemporal correlation.In this way，the problem of traffic velocity missing data recovery can be modeled effectively as a sparse vector recovery one.Finally，based on a largescale dataset，we verify the effectiveness of the proposed algorithm.The experimental results show that the proposed algorithm can accurately recover the missing data when the level of data missing is greater than 50%，and its performance is better than those of other similar algorithms.
2021, Vol. 42 
