The aim of this study was to evaluate the impact of incorporating multispectral data acquired from an unmanned aerial vehicle (UAV) as an extension of standard RGB imagery on the accuracy of road surface material classification using machine learning algorithms. The research was conducted on six road sections differing in pavement type and technical condition. The performance of three ensemble learning algorithms: Random Forest, XGBoost, and LightGBM was investigated. Three input data configurations were considered: RGB imagery, a seven-band dataset comprising RGB channels and four multispectral bands (MS7), and an extended dataset (MS9) additionally including the NDVI and NDRE vegetation indices. For the best-performing model, the influence of texture features calculated within a moving 5x5 pixel window and edge information derived using the Sobel operator was also evaluated. The results demonstrated that the inclusion of multispectral data significantly improved classification performance compared with RGB imagery alone. The highest accuracy was achieved by the LightGBM model using the MS9 dataset, reaching an Overall Accuracy (OA) of 81.31%. The incorporation of texture features further increased the classification accuracy to 93.30% OA. The findings confirm the positive contribution of UAV-acquired multispectral data to the effectiveness of road surface material classification.