The recognition of road surface materials has significant implications for applications like enhanced navigation, traction and stability, predictive maintenance, safety considerations, transportation and infrastructure management, and autonomous driving. In this paper, we aim to accurately identify various materials used in road surfaces, including asphalt, bricks, cobblestone, gravel, among others. To this end, we collected a comprehensive image dataset acquired from dashboard cameras. Each image is annotated with a corresponding surface material groundtruth. Following the data collection, we employed diffusion methods to augment the training data for all surface material classes. Then, we propose a segmentation-classification framework which isolates the road surfaces from surrounding contexts such as buildings, vehicles, and pedestrians. Next, we introduce the road surface sample extraction from the segmentation results. We conducted experiments with various deep-learning models. The experimental results demonstrate that our proposed framework can recognize road surface materials with a high accuracy rate.