Unsupervised Clustering of Public Transport Vehicles Using Deep Morphological Features and Color Histograms
The study developed an unsupervised method to group public transport vehicles into categories like transit routes using both shape and color information. It combines features from a frozen Vision Transformer for structure with color histograms from the HSV space, fusing them with a tunable weight. This approach successfully identified vehicle groups, showing that balancing structural and color data improves classification accuracy.