 This study proposes a framework that combines multi-scale semantic segmentation using CNN and SVM classifier with low-altitude UAV multispectral images to accurately detect asphalt pavement aging states and damaged objects, achieving the highest overall accuracy and recall rate in classification of asphalt pavement images. This article was authored by Yifan Pan, Xiao Chen, Kwam Sun, and others.