Part-based character recognition

在本研究项目中，提出了基于零件的各种字符识别方法。与以往的字符识别试验不同，它可以应用于不同的字符识别任务，如手写字符、场景字符、降级文档图像等。实验结果证明，在手写体字符识别中，各种字体在场景字符和文档图像识别中都有很好的应用。在退化测试集上的实验也证明了基于部分的方法对字符图像的退化具有鲁棒性。此外，基于零件的方法可以应用于字符串和文档图像的识别，而不需要任何特定的分割过程。综上所述，基于零件的方法是解决字符识别问题的较好方法。

In this research project, part-based methods for various character recognition have been proposed. Different from previous trials of character recognition, it can be applied to different character recognition tasks, such as handwritten characters, scene characters, degraded document images, etc. The experimental results proved that in handwritten character recognition, the various fonts in scene character and document image recognition. The part-based methods are also robust against the degradation of the character images, which is also proved by the experiments on degraded test sets. Moreover, the part-based method can be applied to character string and document image recognition without any specific segmentation process. In summary, the part-based methods are the good solutions for the problems in character recognition.