Estimation of Vehicle Usage Rate Based on Capture-Recapture Model with License Plate Recognition Data

This paper presents a modified model derived from the capture-recapture model that is widely used in biostatistics but seldom applied in transportation to estimate the daily utilized vehicles on streets. An estimation algorithm of vehicle usage rate that utilizes a simulative capture-recapture procedure is designed for evaluating traffic situation using license plate recognition (LPR) data. Parameter sensitivity is also analyzed before using real data collected by the LPR systems and the results demonstrate that the model is of high accuracy (average relative error is less than 1%). This paper extends the methodology and algorithm using capture-recapture model to evaluate the running vehicle number on the whole road network by the data collected from LRP systems covering main roads within the fifth ring road in Beijing. The application is helpful for urban transportation managers to understand the recent traffic condition of the whole city, make decisions in time and constitute policies to reduce traffic congestion.