Recognition of speed limit signs in night scene images in Japan

A method to recognize speed limit signs in night scene images is proposed in this paper. The proposed method involves two main stages. First, sign candidates are detected by using the outline of white domain on the signs. This region has little influence on the change of color information or size. Next, the feature and common parts in speed limit signs are distinguished on the basis of their layout. We extract a suitable label from sign candidates and recognize it as a sign through a matching experiment. Experimental results for 596 images show that the proposed method is able to recognize speed limit signs with an accuracy of 94.3%. It is also shown that 423 images without speed limit signs are perfectly distinguished as images having no speed limit signs. © 2013 Institute of Electrical Engineers of Japan. Published by John Wiley & Sons, Inc.