关键词: 道路工程； 沥青路面；三维纹理；小波变换；分形特征；多尺度；抗滑性能

Abstract: The three-dimensional texture properties of asphalt pavement have a decisive influence on the evaluation of the skid resistance of roads. Existing studies mainly focus on the statistical geometric properties of the pavement texture. Although some studies have explored the self-affine similarity and spectral properties of pavement texture, its intrinsic correlation mechanism with the skid resistance of pavements still needs to be further investigated. Therefore, a multi-resolution fractal analysis method was proposed for multi-scale analysis of pavement texture. Firstly, by integrating wavelet transform and fractal theory, the wavelet decomposition was performed on the textures of three typical graded specimens under different sand accumulation states. Then, the high and low frequency textures were reconstructed layer by layer, and the fractal and spectral features were extracted to construct a multi-resolution fractal data set. Finally, the effectiveness and robustness of the constructed multi-resolution fractal dataset was analyzed and verified by comparing the mainstream machine learning models such as XGBoost, CatBoost, NGBoost and LightGBM. The results show that LightGBM has the highest goodness of fit, with an R2 of 88.90% and a root mean square error of 3.59. In addition, it is revealed that under different sand accumulation conditions, the single fractal dimension D exhibits a variation rule of first increasing and then decreasing as the number of decomposition layers progresses, ultimately tending towards stabilization. Moreover, the multiple fractal characteristics of three different graded samples are furtherly investigated, revealing their differences in multiple fractal spectra.

Key words: highway engineering; asphalt pavement; 3D texture; wavelet transform; fractal features; multiscale; skid resistance