Validating traffic models using large-scale automatic number plate recognition (ANPR) data

Abstract

The development of reliable strategic traffic models relies on comprehensive and accurate data, but traditional survey methods are time-consuming and expensive. Manual surveys often yield small samples that require estimated expansion factors to enable the data to represent the population. Modellers have turned to new data sourced from various electronic devices to improve the reliability of the data. Automatic Number Plate Recognition (ANPR) data is one such data source that can be used to extract travel time, speed and partial origin-destination (OD) information. This study assesses ANPR data in terms of its comprehensiveness and accuracy, and shows how it can be used for the validation of strategic traffic models. Data was obtained from the Gauteng freeway system’s Open Road Tolling (ORT) gantries for a period of several months. A new methodology is developed to process traffic model outputs such that they are directly comparable to the partial origin-destination outputs derived from the ANPR data. It is shown that comparing the model distribution against observed ANPR data highlights potential trip distribution issues that are not detected using standard model validation techniques.