Abstract
Ministry of Land, Infrastructure, Transport and Tourism of Japan is going to make maximum use of vehicle detectors installed at national roads around the country and efficiently gather traffic volume data from wide areas by estimating traffic volumes within adjacent road sections based on the constant observation data obtained from the vehicle detectors. Efficient processing of outliers and missing values in constant observation data are needed in this process.
Focusing on the processing of singular and missing values, the authors have developed a series of algorithms to calculate hourly traffic volumes in which a required accuracy is secured based on measurement data obtained from vehicle detectors. The algorithms have been put to practical uses. The main characteristic of these algorithms is that they use data accumulated in the past as well as data from constant observation devices in adjacent road sections. This paper describes the contents of the developed algorithms and clarifies their accuracy using actual observation data and by making comparis on with other methods.
