Recognition of vehicle registration mark on moving vehicles in an outdoor environment

We propose a method for segmenting vehicle registration plates and recognizing the registration mark on a moving vehicle in an outdoor environment. The algorithm first segments the plate from the vehicle and other complex objects in view, based on the plate color and its dimension. The segmented plate is then corrected in orientation and size, before being matched with a sub-set of templates in the database. The matching adopts a highly tolerant scheme allowing characters to have certain degree of shifting, rotation and mis-match. This proves to be an important criterion for successful recognition. Preliminary tests show that this method offers high success rate and high confidence level.