Reliable detection and skew correction method of license plate for PTZ camera-based license plate recognition system

The skewness of license plates affects negatively all processing steps of license plate recognition including character segmentation and character classification. The skewness of the license plate in the captured image may appear severely under PTZ camera environments since license plate appears to have various postures in the captured image due to panning and tilting positions of PTZ cameras. Thus far, most of previous works on this subject have mainly resolved skewness problem due to rotation in plane, but not rotation in depth. In this paper, by utilizing planar projective transformation (planar homography), we propose a reliable skew correction method of license plates which handles also the skewness due to rotation in depth. For more reliable deskewing process, the license plate needs to be detected more reliably so that this paper also proposes a reliable license plate detection method. The effectiveness of the proposed skew correction method as well as the proposed license plate detection method is verified though experiments.