License plate recognition using MSER and HOG based on ELM

In this paper, an effective method for automatic license plate recognition (ALPR) is proposed, on the basis of extreme learning machine (ELM). Firstly, morphological Top-Hat filtering operator is applied to do the image pre-processing. Then candidate character regions are extracted by means of maximally stable extremal region (MSER) detector. Thirdly, most of the noise character regions are removed according to the geometrical relationship of characters in standard license plates. Finally, the histograms of oriented gradients (HOG) features are extracted from each character of every plate detected and the characters are recognized by the classifier trained though the ELM. Experimental evaluation shows that our approach significantly performs well in the ALPR systems.