Character segmentation for license plate by the separator symbol´s frame of reference

Character segmentation is the middle stage of the license plate recognition algorithm. The precision of the character segmentation is blocked by two factors. One is the character area always have errors for the reason of license plate location algorithm, which made the real margin of the license plate define impossible. The other is that the characters themselves are illegible, or the strokes are broken or conjoint, namely the image is degraded. So characters´ borders are also very difficult to obtain. To solve these problems, a character segmentation method bases on the separator symbol´s frame of reference is proposed in this paper. At first, figuring out the separator symbol´s region by using the uniqueness of projection feature of separator symbol´s region, then constructing the characters´ borders of pre´ segmentation based on the separator symbol´s region combines some prior knowledge of license plate format. After analyzing the projection feature on character area, each character can be segmented accurately. Experiment results show that this method can segment license plate characters precisely and robustly.