基于Laplacian扩展算子的车牌字符分割方法研究

Research on the Segmentation of License Plate Character Based on Laplacian Extension Operator

在车牌识别技术中,车牌图像的二值化处理是字符分割的关键技术。分析了传统基于灰度二值化方法在特殊场景中应用的缺陷,提出了一种基于Laplacian扩展算子的二值化方法,该方法首先通过计算出车牌图像的边缘灰度分布特征,然后进行自适应阈值三值化处理,获取黑、白、灰三值图,再结合车牌颜色信息,准确判定字符与背景,实现图像二值化。对于存在阴影覆盖和局部过曝光等现象的车牌图像,该方法能够充分保留笔画的完整性,显著提高车牌图像的二值化效果。

Binarization processing has always been the basis of character segmentation in the technique of LPR.With the limitation in traditional applications of grayscale binarization methods in peculiar scenes,a new binarization application based on extended Laplacian operator is presented in this paper.First the marginal gray scale distribution feature of the license plate image is calculated so as to continue a three-valued image processing based on self-adaptive threshold to acquire a ternary image of black,white and gray,then the characters and background can be determined accurately according to the color of the licence plate,finally the ternary image is converted into a binary image.The proposed method is particularly advantageous in maintaining the integrity of strokes when the license plate image is covered by shadows or has overexposed areas,and significantly improves the effectiveness of the binarization processing.