Accurate knowledge of surface roughness is essential for understanding key phenomena such as tire-road friction and wear rate. However, high-resolution surface characterization typically requires scanning devices that are time-consuming to use and expensive. In time-critical contexts like motorsport, where setup windows are narrow, having a faster and more accessible alternative could provide a significant advantage. This paper presents a fast, cost-effective and physically interpretable methodology to reconstruct the three-dimensional roughness of road surfaces directly from grayscale images. The results demonstrate a strong correlation between 2D photometric information and 3D surface features, preserving both the height distribution and spectral content. The proposed framework combines light-intensity features extracted from the image with a Gaussian Process Regression (GPR) model, offering high physical controllability and reduced data requirements compared to deep learning approaches such as convolutional neural networks. The model was validated under various conditions and proved effective even on unconventional textures, including sandpaper and marble samples. It also showed robustness to different image formats. Overall, the method provides a practical and innovative alternative for surface roughness assessment, with strong potential for applications in vehicle dynamics, tire modeling and tribology.
原文链接
分享
引用
