一种预测纸币序列号先知的字符分割方法

纸币在流通过程中不可避免地引入各种污染, 会影响纸币序列号的正常分割, 甚至引发分割失败。为了提高分割结果的准确性, 提出了一种预测先验知识的字符分割方法。对于某种面值的纸币, 根据其特有的号码排列方式定义马尔可夫链的状态, 通过求得马尔可夫链的前向识别, 可以预测纸币序列号的先验知识, 再配合连通区域法和垂直投影法, 最终得到序列号的最优分割位置。在各面值的人民币以及部分外币图像组成的数据集上测试该方法可行性, 实验效果与传统的投影法和区域连通法对比显示了该分割方法的准确性。该方法对于有污痕、拆痕、字符粘连等情况的图像能进行准确分割, 比传统分割法具有更好的分割性能。

Segmentation of banknote serial number with predicting priori knowledge

As the banknote character segmentation results are seriously influenced by all kinds of noise and pollution, this paper proposed a character segmentation algorithm based on predicting priori knowledge. For some kinds of banknotes, it used the unique number arrangement to define the state of Markov chains, and then calculated the set of Markov transitions to get its priori knowledge. Combined with connected-component-based and projection-based algorithms, it obtained the optimized segmentation result. Tests in both various denominations of RMB and parts of the foreign currency data set show that this method is feasible, the results comparisons with the traditional connected-component-based and projection-based algorithms display that this method is more accurate. For those images like dirty, folding-traces, character-adhesion etc, this method can get an optimal segmentation result and has better performance than the traditional algorithms.