Abstract: The local and global characteristic of characters of the license plate is fully used and the concept of the similarity between a character string and the license plate is introduced, on the basis of which a new algorithm for the character segmentation of the license plate is proposed. Using connected-component searching and projection, this algorithm enumerates all the possible segmentations. And then it analyses the segmented characters using clustering. Finally, it selects a character string that has the maximal similarity to the license plate to segment the character. Experimental results show that the algorithm has good feasibility and effectiveness.

Key words: license plate recognition, character segmentation, projection, cluster