Exploring Textural Characteristics of Spatiotemporal Traffic Contour Maps

Abstract

This study presents an approach to extract properties from spatiotemporal traffic speed contour maps using tools from the field of digital image analysis. The new measures are derived from second-order statistics and quantify properties such as smoothness, homogeneity, regularity, and randomness in traffic operations. Four measures were selected: angular second moment (ASM), contrast (CON), inverse difference moment, and entropy. Each measure was used to characterize speed contour maps derived from different traffic conditions. To avoid information redundancy, correlation was examined between first- and second-order measures, which resulted in disqualifying one of the four measures. The sensitivity of the new measures to variations in traffic conditions was also investigated using nearly 14,000 30 min speed contour maps generated from a section of 5.44 km of the freeway in five weekdays. It was found that the speed range of 32–48 km/h exhibited the highest randomness (entropy) and least smoothness (ASM). The lowest level of homogeneity (CON) was observed in the range of 48–80 km/h. The new measures were also used to evaluate the level of service.