Improved prediction of pavement friction considering multi-directional friction and texture indexes

Determination of pavement friction from digital images of pavement surface texture presents an effective alternative to physical friction measurements. Currently, virtually all texture index-based pavement friction prediction methods ignore the fact that both pavement friction and texture characteristics vary with test direction. Ignoring the directionality of pavement friction and texture inevitably introduces errors in pavement friction prediction. This study proposes a procedure that evaluates pavement texture indexes in any given direction, and develops statistical relationships to predict pavement friction in the given direction accordingly. Field measurements of pavement texture and friction were made along two asphalt pavement curves and three straight asphalt pavement sections to demonstrate that there exist statistically significant relationships between directional pavement friction and the corresponding directional pavement texture characteristics. Ten directional texture indexes were identified based on texture characteristics of amplitude, shape, roughness, and distribution of peaks. All ten indexes were found to have positive correlations with the corresponding directional pavement friction measured in terms of the British pendulum number (BPN). A statistically significant regression equation with R 2 of 0.813, with contributions from texture indexes related to geometric characteristics of texture peaks, were identified for the purpose of pavement friction prediction. Test results demonstrated that the new approach, which accounts for the directionality of friction and texture, produced markedly better pavement friction predictions than the conventional method.