Research on Adaptive Preprocessing License Plate Location

A robust algorithm for license plate location is proposed, on the basis of connected edge analysis in conjunction with a novel adaptive preprocessing technique. Firstly, with non-conglutination feature of connected edge produced by license plate characters in binarized vertical edge image, filtering and row-scanning location are conducted. Adaptive preprocessing function is then introduced. An input image is regarded as the preprocessing-needed one if non-preprocessing location fails. And preprocessing relocation for this sort of images is carried out. The algorithm is tested with 388 natural-scene vehicle images of different backgrounds and ambient illumination, and 357 of them are located properly (92.0%). Experimental results show that it is advantageous of the algorithm to achieve higher success rate and more robust real-time performance, comparing to full-preprocessing location algorithm.