Abstract: In the comment information of aerial photos，there exist important imaging parameters.The comment information must be automatically recognized for later use.In this paper，the process of comment information is introduced，including character region location，binarization，character segmentation and character recognition based on neural network.Preprocessing is lucubrated，especially binarization of comment information block.A thresholding method is proposed.The experiments demonstrate that the system implemented in this paper has a high recognition rate in general，and that the thresholding method proposed is comparable to both Sauvola algorithm and White algorithm while the execution time is greatly reduced.

Key words: character recognition, thresholding, character segmentation, neural network

摘要： 航拍胶片中的注释信息记录着重要的成像参数。胶片数字化后需要自动识别出注释信息供后续处理使用。介绍了注释信息字符识别的主要步骤，包括字符区域定位、二值化、字符分割和基于神经网络的字符识别；重点研究了预处理过程，特别是注释信息块的二值化，提出了一种新的阈值方法。大量的实验表明，实现的系统总体上具有很高的识别率，该阈值方法的效果能与Sauvola算法和White算法相比，而所需时间却大大减少了。

关键词: 字符识别, 阈值, 字符分割, 神经网络