一种基于反馈的车牌图像二值化方法

A Method of License Plate Image Binarization Based on Feedback

这篇论文提出了一种基于字符识别可信度反馈的车牌图像二值化方法 ,该方法先通过二值化阈值对车牌区域图像二值化,车牌字符分割后,再通过快速字符识别方法得到车牌的识别可信度,利用该可信度指导二值化阈值的选取,形成反馈机制,从而提高了车牌字符分割的准确性,减轻了利用深度学习进行车牌字符分割训练过程中车牌字符位置样本标注的工作量。另外,为了提高计算效率,该论文还提出了相应的快速简化算法以及一种通过车牌区域背景与前景像素数之比确定二值化阈值的方法。

This paper introduces a image binarization method based on feedback of the character recognition reliability of license plate, through the license plate image binarization with a threshold and license plate character segmentation, the recognition reliability of license plate is obtained by a fast character recognition method, the method uses the recognition reliability to choose the best binarization threshold. With this kind of feedback mechanism, the method improves the accuracy of the license plate character segmentation and reduces the workload of labeling license plate character position of training samples in the use of deep learning for license plate character segmentation. In addition, in order to improve the computational efficiency, this paper also proposes a corresponding fast simplification algorithm and a method to determine the binarization threshold by the ratio of the background and the foreground of the license plate image.