Adaptive License Plate Extraction Method

Recently most license plate extraction algorithms are proposed under some certain conditions. This paper proposes a license plate extraction method that can be self-adaptive to images with various sizes under different illumination conditions. First, a morphology based image preprocess is applied for background elimination. Then we get the level clipping by a self-adaptive binary thresholding the image vertical projection. After this, a morphology closing operation following a opening operation is implemented,then connected component analysis(CCA) is carried out to detect plate candidate regions,which will be initially verified by minimum width,height,aspect ratio and Euler number.The real license plate will be obtained by a novel license plate characters uniform measure after character segmentation.Experimental results show our algorithm is widely suitable for license plate extraction in images that of different sizes and different illumination.