Railway sign detection and classification

This paper illustrates an application of computer vision at Italian railway signalling. From video frames captured on a driving train, each image is processed for robust object detection based on HSI color model, threshold analysis. Next, a classification algorithm based on shape feature and template matching is performed to detect the relevant railway signs and to discard objects of railway infrastructure. Our algorithm works with minimal computation and simple image processing techniques. Our approach can detect sign candidates in presence of complex background. Experimental results show relevant percentage of accuracy on a standard PC