Publicado en International Conference on Systems, Signals, and Image Processing Abstract: This paper describes an implemented solution to automatically detect and recognize in real-time both speed limit warning signs and speed limit signs in railway travel videos recorded from the driver's cab. The lack of available high-quality videos to train this kind of systems and the involved complexity of the rail scenes, with non-controlled illumination conditions, make it challenging the considered problem. Our framework achieved interesting recognition results of around 95% for both signs types and digits recognition.