中文摘要：

针对纸币冠字号识别中传统方法速度慢和准确率低的问题，改进预处理的关键算法，提出一种组合特征识别方法。通过改进图像定位、旋转校正、二值化和去噪滤波等预处理算法，提高处理速度和二值图像质量；在此基础上，提出一种结合必选特征和可选特征的组合特征提取方案，采用多叉树分类器设计组合特征识别算法。实验结果表明，与传统的定位、旋转、二值化和字符识别算法相比，该方法具有更高的识别率和处理速度。

英文摘要：

To solve the problem of low accuracy and low speed of the traditional methods in paper currency number recognition, the key algorithms of preprocessing were improved and a method of combined-feature recognition was proposed. The processing speed and the quality of binary image were improved by modifying the preprocessing algorithms, including the image location, the rotation, the binarization and the filtering. Based on these, a combined-feature extraction solution combining the required fea- tures and optional features was proposed, and a combined-feature recognition algorithm using multi-tree classifier was designed. Experimental results reveal that, compared with the traditional location algorithm, rotation algorithm, binarization algorithm and character recognition algorithm, the presented method has higher recognition accuracy and processing speed.