Automatic Number Plate Recognition (ANPR) data is increasingly being used to characterise travel time between two points on the road network. However, it is important to filter out unwanted travel time records from the ANPR dataset. Previous research has shown the overtaking rule to outperform other ANPR data cleaning filters. However, this filter suffers from poor performance at very low flows (typically less than 60 vehicles per hour). This paper proposes two extensions of the overtaking rule; one to account for low flows in all conditions, and the other to account for low flows in only off-peak conditions. Using data from the ANPR system in central London it is shown that either extension overcomes the failing of the standard overtaking rule at low flows. For the covering abstract see ITRD E134653.
Abstract