Pavement macrotexture estimation using principal component analysis of tire/road noise

An investigation on the prediction of macrotexture Mean Texture Depth (MTD) of pavement from a moving vehicle is conducted. The MTD was predicted by using the tire/road noise measured from a microphone mounted underneath a moving vehicle. Principal Component Analysis (PCA) is used to filter noise from microphone data prioer to estimating its energy over an optimally selected bandwidth. Energy obtained using this method is named PCA energy, hence the developed method for MTD estimation is termed as PCA Energy Method. The acoustic energy is assumed to have positive linear correlation with MTD of pavement. Moreover, PCA was used to differentiate important information about the road surface from noisy data while vehicle is moving, yielding a set of principal component vectors representing the conditions of each road section. This principal component vector was used to compute the PCA energy that is to be used for MTD prediction. The frequency band most relative to pavement macrotexture was determined to be 140 to 700 Hz through theoretical and statistical research. Then, a MTD prediction model was built based on a Taylor series expansion with two variables, PCA energy and the driving speed of the vehicle. The model parameters were obtained from an engineered track (interstate highway) with known MTD, and then applied to urban roads for the feasibility test. The accuracy of the model is 83.61% for the engineered track, which is 10% higher than the previous energy-based methods without PCA treatment. Moreover, applicability of the model is increased by the extended MTD prediction range between 0.2 and 3 mm compared to that of the engineered track having 0.4 to 1.5 mm. In addition, the MTD could be predicted every 7.8 meters and with good repeatability in the urban road test, which proves the feasibility of the proposed approach. Therefore, the PCA Energy Method is a reliable, efficient, and cost effective way to predict MTD for engineering applications as an important index for pavement friction prediction and pavement segregation identification.