Adaptive Spike Removal Method for High-Speed Pavement Macrotexture Measurements by Controlling the False Discovery Rate

Abstract

Tire–pavement interactions, such as friction, tire–pavement noise, splash and spray, and rolling resistance, are significantly influenced by pavement macrotexture. Accurate texture data collection and analysis at a network level are key to achieving the desired level of safety, comfort, and sustainability of pavements. This study addressed the problem of noise in the form of spikes revealed in dynamic measurements that were performed with the use of vehicle-mounted lasers to measure macro-texture at traffic speed. The presence of spikes in collected data leads to erroneous texture measurements that do not reflect the actual pavement texture profile. As a solution to this problem, an innovative denoising methodology was developed. It consisted of an algorithm that determined the distribution of texture measurements with the use of the family of generalized Gaussian distributions, which allowed for the tail of the distribution to be heavier or thinner than the normal distribution, and with the use of the false discovery rate (FDR) method that controlled the proportion of wrongly identified spikes to all identified spikes. The FDR control allowed for an adaptive threshold selection that differentiated between valid measurements and spikes. Finally, the validation of the method showed that the mean profile depth (MPD) results obtained with denoised dynamic measurements were comparable to MPD results from the control devices on all the pavement sections investigated, making this method a significant step in the development of standardized procedures that use these devices for texture investigation at the network level.