Hybrid Method for License Plate Detection from Natural Scene Images

摘要

License plate detection is a key part in vehicle license plate recognition system. In this paper, we present a hybrid method for license plate detection from natural scene images for the all-day traffic surveillance environment. The proposed method includes two stages: rough detection and accurate detection. Coarse detection stage based on color edge and morphology can help finding the region of interest quickly; Accurate detection stage based on HOG and SVM accurately detect the vehicle license plate. The effectiveness of the proposed method has been proven by the experimental results on a large database of images.