基于细胞神经网络和网格特征的碑刻文字识别

Table Recognition Based on CNN and Grid Feature

碑刻作为文字的载体,具有极其重要的历史价值和艺术价值。但是,因为时间的久远,其拓本大量的噪声对碑刻文字的辨识产生一定的影响。针对这种情况,提出了一种新方法,首先利用细胞神经网络(CNN)技术,去除单个碑刻文字的噪声,然后用网格特征提取方法对其进行识别。本文在MATLAB平台上进行了实例模拟,取得了较好的效果。

As a carrier of the text,the table has very important historical and artistic value.However,because of a long time,a lot of noises on the rubbings have impact on recognizing text.This paper proposes a new method for this case.Firstly,remove the noises of a single table using CNN technology.And then recognize the table using the grid feature extraction method.The simulation example is based on MATLAB software,and we achieve good results.